security-agent-skills plugin
White-box bug hunting, AI-agent and LLM red-teaming, cloud identity and CI/CD trust, client-app trust surfaces across browser and editor extensions and Electron, wire-protocol and token trust across gRPC, WebSocket, and JWT, enterprise identity lifecycle across SCIM and JIT provisioning, directory sync, MFA and key lifecycle, and Kerberos delegation, infrastructure-as-code and container hardening, mobile app security, smart-contract and DeFi review across EVM and Move, skill and MCP supply-chain vetting, API and object-authorization depth, host privilege escalation, network-service and transport trust, firmware and embedded-software trust, web trust boundaries, authentication and session depth, rate-limiting and multi-tenant isolation, injection and resource-exhaustion depth, file-upload and content handling, defensive detection and logging, supply-chain and dependency risk, business-logic and access-control depth, the classic web-attack classes across XSS, template and command injecti
How to install the security-agent-skills plugin
Run both commands inside Claude Code. The first adds the marketplace once; the second installs the plugin.
/plugin marketplace add UnboundCompute/security-agent-skills
/plugin install security-agent-skills@unboundcomputeWhat's inside security-agent-skills
194 parts read from the plugin's own files.
Skills 194
Claude loads it when the task matches- adjudicating-dependency-cve-reachability
Decide whether a CVE in a dependency actually exposes your application before you scramble to patch: is the vulnerable function on a real call path from your code, do the trigger preconditions hold, and can an attacker control the input that reaches it. Use when an advisory, SCA scan, or dependency bot flags a CVE and you must separate a genuine exposure from unreachable noise, or justify why you are or are not affected. Covers pinning the vulnerable symbol, call-path reachability, precondition checks, and taint from your entry points.
- adjudicating-taint-paths
Decide whether a whitebox lead is a real bug by tracing taint from an untrusted source to a dangerous sink and confirming every hop against live source. Use after a scanner, a candidate list, or your own reading surfaces a "this looks dangerous" sink (SQL exec, system/exec, file open, deserialize, template render, redirect target, memcpy) and you must decide whether attacker- controlled input actually reaches it - or kill the lead with evidence. Covers forward and reverse taint, witness paths, sanitizer analysis, and the evidence rules that separate a finding from a false positive.
- auditing-account-abstraction-and-paymaster-trust
Audit an ERC-4337 account-abstraction deployment for trust misplaced in the user-operation lifecycle: a smart account whose validation accepts a signature or nonce it should reject, a paymaster that agrees to sponsor gas for operations it should not so an attacker drains its deposit, a bundler or entry-point assumption that lets a user operation be replayed or reordered for gain, and validation logic that reads mutable state or reaches outside its allowed scope. Covers smart-contract wallets, paymasters, bundlers, and the entry point in an account-abstraction stack where a user operation is validated and sponsored before it executes. Use when a user operation is validated, paid for, and executed by separate parties and that trust split is the boundary. The crafted user operation is the source, the drained paymaster or unauthorized execution is the sink, and the over-permissive validation or sponsorship rule is the bug.
- auditing-account-recovery-and-reset-trust
Audit password reset and account recovery flows for the trust that lets an attacker take over an account: a reset token that is guessable, long-lived, reusable, or not bound to the account it was issued for, a recovery path that verifies a weaker factor than login and bypasses multi-factor, a reset link whose host comes from an attacker-controllable header so the token leaks, and a recovery that trusts an unverified email or phone change to redirect the reset. Covers the recovery surface of authentication systems, where resetting a credential or recovering access is the alternate door into an account. Use when an application offers password reset or account recovery and that flow is a path to authentication. The attacker-driven recovery request is the source, the account it takes over is the sink, and the weak token, bypassed factor, or leaked reset link that grants it is the bug.
- auditing-admission-control-policy-gaps
Audit cluster admission control for gaps that let a non-compliant or hostile workload through: a validating webhook that fails open when its backend is unreachable, a policy that scopes by namespace or label and misses the namespaces that matter, a mutating webhook whose changes are trusted downstream, and an ordering or exemption that lets a privileged pod bypass the checks meant to stop it. Covers Kubernetes admission webhooks and policy engines that are supposed to enforce workload security at create and update time. Use when admission policies are the control that keeps privileged, unsigned, or over-permissioned workloads out of the cluster. The workload submitted for admission is the source, the admitted resource is the sink, and the policy gap that let it through is the bug.
- auditing-ai-agent-permissions
Audit what an AI agent is actually allowed to do versus what its task needs. Covers excessive agency (tools, scopes, and autonomy beyond the job), missing human-in-the-loop gates on irreversible actions, over-broad credentials and their blast radius, sandbox and code-interpreter escape, unfiltered egress, and unbounded resource or spend (denial-of-wallet). Use when granting an agent a tool or scope, reviewing an agent's permission posture, or deciding which actions need approval. The model's restraint is not a control; permissions are.
- auditing-android-component-exposure
Audit an Android app for components another app on the device can reach and drive, after the manifest export flags and permission gates are resolved. Covers an activity, service, broadcast receiver, or content provider exported without a permission gate, an intent filter that makes a component implicitly exported, a provider that grants URI access or exposes files across the app boundary, a permission declared with a weak protection level, and a component that trusts intent extras from an untrusted caller for a sensitive action. Use when reviewing the manifest and the component code that handles an inbound intent, not the deep-link URL trust that a WebView loads (that is the deep-link skill). A cross-app caller is the source, a reachable exported component acting on the intent is the sink, and a sensitive component another app can invoke unguarded is the bug.
- auditing-android-intent-redirection-and-pendingintent
Audit Android privilege and access leaks through intent redirection and mutable pending intents, where a privileged component receives an intent carrying a nested intent it then launches, or hands out a pending intent an untrusted app can fill in, so the untrusted caller reaches a protected component or acts with the privileged app's identity, because the component forwards an attacker-supplied intent without constraint or the pending intent is mutable and under-specified. Use when an Android app forwards intents extracted from input or shares pending intents with other apps. Covers nested-intent redirection to protected components, mutable pending intents, implicit pending-intent delivery, and forwarding that carries the app's permissions. The attacker-supplied nested or fillable intent is the source, the privileged component launching or sending it is the sink, and reaching a protected target with the app's identity is the bug.
- auditing-ansible-become-and-vault-trust
Audit configuration-management privilege escalation and secret handling for trust that runs as root on every managed host: a task that escalates with become across a whole play when only one step needs it, a role or variable sourced from an untrusted place that runs under that escalation, a vault-encrypted secret whose decryption key is exposed to the runner or logged, and a templated value or module argument that takes attacker-influenceable input while privileged. Covers Ansible and similar agentless tools where a control node runs plays that escalate privilege and decrypt secrets across a fleet. Use when playbooks escalate with become or handle vault secrets across managed hosts. The untrusted role, variable, or input running under escalation is the source, the privileged task or decrypted secret is the sink, and the over-broad escalation or exposed key is the bug.
- auditing-api-key-and-token-lifecycle
Audit the lifecycle of API keys and access tokens for weaknesses that let one keep working past its intended bounds: a key issued with broader scope than the caller needs so a leak grants far more than one function, a key or token with no expiry that stays valid indefinitely, a revocation path that does not actually stop the key so a rotated or compromised credential keeps authenticating, a key that leaks into logs, URLs, client-side code, or error messages and is never rotated, and a token whose scope or audience is not enforced on use so it works against endpoints it was never meant for. Use when a long-lived programmatic credential authenticates a caller and the bounds on that credential (scope, expiry, revocability) are the boundary. The over-scoped, unexpiring, or leaked key is the source, the access it grants past its intended bounds is the sink, and the missing scope limit, expiry, or working revocation is the bug.
- auditing-ble-and-gatt-authorization
Audit a Bluetooth Low Energy device for missing authorization on its GATT attributes: a characteristic performing a sensitive action or revealing sensitive data readable or writable by any peer, pairing or bonding that is not required or falls back to an unauthenticated Just Works mode with no link encryption, an authorization decision the device pushes to the mobile app instead of enforcing on the peripheral, and a replayable command a sniffer can capture and resend. Covers BLE peripherals, wearables, locks, medical and IoT devices, and their GATT services where a connected peer reads or writes characteristics. Use when a peer can connect over BLE and the peripheral's own enforcement of who may read or write each characteristic is the boundary. The unauthorized connected peer or captured command is the source, the sensitive read, write, or action on the peripheral is the sink, and the missing pairing, characteristic-level authorization, or replay protection is the bug.
- auditing-break-glass-account-trust
Audit emergency break-glass and privileged-access accounts for the ways their standing power outlives the emergency they exist for: a break-glass account with a static, shared, or never-rotated credential that is a permanent superuser login rather than a sealed last resort, an emergency account excluded from the MFA, conditional-access, or logging controls covering other admins so its use is easy and invisible, a just-in-time elevation that grants more than the task needs or is not revoked when the task ends, an emergency-access path with no approval, time bound, or alerting so it is used routinely, and a break-glass use that triggers no review afterward. Use when a standing emergency identity or an elevation path holds power beyond ordinary admins and the controls around when and how it is used are the boundary. The break-glass credential or elevation request is the source, the unbounded or unaudited privileged access is the sink, and the missing rotation, control coverage, or use-time alerting is the bug.
- auditing-browser-extension-trust
Audit a browser extension (Manifest V3) for a trust boundary another web page or extension can cross to reach a privileged capability, after the permission scope and the message-sender checks are resolved. Covers an externally_connectable or onMessageExternal handler that verifies an origin but not the calling script, a content-script-to-background message handler with no sender validation, host permissions broader than the extension needs, a web_accessible_resources page that acts on URL parameters, an injected content script writing page-controlled data to a DOM sink, and a weak or eval-permitting content-security policy. Use when reviewing the manifest, background and content scripts, and cross-context message passing, not the web-page DOM sink taxonomy the client-side DOM skill owns. An untrusted web origin or another extension is the source, a privileged extension API or DOM sink is the sink, and a message reaching it without a sender-and-origin check is the bug.
- auditing-cicd-oidc-trust
Audit continuous-integration pipelines for the trust they extend to untrusted input: workflows that run on incoming change requests from forks while holding repository secrets, steps that let attacker-controlled content reach a privileged command, and cloud role trust conditions that accept a pipeline's short-lived token too broadly. Covers secret and token exposure on fork-triggered runs, poisoned-pipeline execution, and over-broad trust on the identity claim a pipeline presents to a cloud account. Use when reviewing CI/CD configuration, pipeline identity, or the boundary between a build and the cloud it can reach. An exploitable token or command from untrusted input is the finding.
- auditing-clickjacking-and-ui-redressing
Audit a web application for UI-redressing attacks where an attacker frames the real site and tricks a user into acting on it unknowingly: a sensitive page that can be embedded in an attacker's iframe because it lacks frame-ancestors or X-Frame-Options, a state-changing action reachable by a single click that an overlay or transparent frame steers, a drag-and-drop or keystroke redressing that captures input meant for the attacker page, and a confirmation step that a framed overlay hides. Covers web pages with authenticated, state-changing actions (settings changes, purchases, approvals, connect flows) that could be loaded inside a frame the user cannot see. Use when a sensitive action can be triggered by a click and the page can be framed by another origin, making framing the boundary. The attacker page that frames or overlays the real site is the source, the unknowing state-changing click is the sink, and the missing framing protection or unguarded one-click action is the bug.
- auditing-container-image-build-hardening
Audit container image build definitions (Dockerfile, containerfile, and the compose or run config that sets runtime flags) for an image that ships over-privileged or carrying a secret, after multi-stage discards and deploy-time overrides are accounted for. Covers an image that runs as root, a secret baked into a layer, remote content pulled unpinned or unverified, a mutable or untagged base, an over-broad copy that pulls in local secrets and history, and a dangerous runtime request such as privileged mode or a sensitive host mount. Use when reviewing the image build plane, not the deploy-time security context or the cloud resource definition. The build definition is the source, the shipped image or run config is the sink, and privilege or a secret that survives into the final image is the bug.
- auditing-container-image-provenance
Audit how a cluster decides which container images to trust and run: an image referenced by a mutable tag rather than a content digest, a workload pulling from a registry that admits unsigned or unverified images, a signature or attestation policy that is configured but not enforced at admission, and a base image or layer whose origin the pipeline never verified. Covers Kubernetes and container platforms where the image a workload runs is the code that runs, and where tag mutability, signing, and provenance decide whether it is the intended artifact. Use when workloads pull images whose signing and provenance are not enforced end to end. The unverified image reference is the source, the running container is the sink, and the code that runs without proven provenance is the bug.
- auditing-container-runtime-and-socket-exposure
Audit whether a workload can reach the container runtime and thereby control the host: the container runtime socket mounted into a pod or bound into a build or CI container, a privileged sidecar that talks to the runtime to launch or inspect containers, a runtime API exposed over a reachable port, and tooling that needs the runtime and is given it far more broadly than the one operation requires. Covers container hosts where reaching the runtime socket or API grants the ability to start privileged containers, mount the host filesystem, and take over the node. Use when a workload, build, or agent is given access to the container runtime. The workload with runtime access is the source, the container runtime is the sink, and the socket or API exposure that grants host control is the bug.
- auditing-cors-and-cross-origin-trust
Audit the code and configuration that decide cross-origin access, for trust a browser turns into a read of authenticated data. Covers a response that reflects an arbitrary request origin into the allow-origin header alongside allow-credentials, an allowlist that accepts the null origin, allowlist matching by prefix, suffix, substring, or an unanchored pattern that a lookalike origin satisfies, the origin header trusted as an authorization or request-forgery defense, and a cross-window message handler that acts on data without an exact origin and source check. Scoped to the code and config that build the decision, not a live-header scan. Use when reviewing cross-origin response headers, origin-based access logic, or cross-window message handlers. The request origin or the posted message is the source, the credentialed cross-origin read or the message sink is the sink, and trusting an attacker-set origin is the bug.
- auditing-cross-account-role-trust-boundaries
Audit cross-account IAM role assumption for trust policies that let the wrong principal assume a role: a trust policy with a wildcard or overbroad principal, a missing or unverifiable external ID on a third-party role, a confused-deputy path where a vendor assumes your role on any customer's behalf, and a role chain that reaches privileges the origin principal should never hold. Covers AWS assume-role trust policies, condition keys that should scope who may assume, and the transitive reach of one assumption into the next. Use when roles in one account can be assumed from another account, a partner, or a service, and the trust policy is the boundary. The external principal permitted by the trust policy is the source, the assume-role grant is the sink, and the trust scope wider than the intended relationship is the bug.
- auditing-cross-chain-bridge-and-message-trust
Audit a cross-chain bridge or messaging protocol for misplaced trust in messages that cross chains: a destination contract that accepts a mint or release on a forged or unverified proof of a source-chain event, a message whose signer set or validator quorum can be spoofed or is too small, a message that can be replayed on the destination or across chains for a repeated withdrawal, and a lock-and-mint or burn-and-release accounting that a crafted message pushes out of balance. Covers token bridges, message-passing layers, and any protocol where an action on one chain is authorized by an event claimed to have happened on another. Use when a destination-chain action depends on trusting a source-chain event and that verification is the boundary. The forged or replayed cross-chain message is the source, the unauthorized mint, release, or state change is the sink, and the missing or spoofable source-event verification is the bug.
- auditing-csrf-and-unsafe-state-changes
Audit state-changing endpoints for cross-site request forgery, where a request that rides the victim's ambient cookies is authorized on that session alone with no unpredictable, session-bound proof the request came from the app, letting an attacker page trigger the change as the victim. Use when reviewing forms and actions that modify data, change settings, move funds, or alter access, and how each is protected. Covers missing or unvalidated tokens, tokens not bound to the session, cookie-only same-site reliance and its gaps, state-changing GET requests, and content-type or method-override assumptions. The cross-site request on the victim's session is the source, the state-changing endpoint is the sink, and acting without unpredictable session-bound proof of origin is the bug.
- auditing-datastore-exposure-and-abuse
Audit in-memory and cache datastores such as Redis and memcached for exposure and command abuse: an instance reachable without authentication, a request that composes datastore commands from untrusted input, or server-side scripting and module or config commands that reach code execution or a file write. Covers unauthenticated network exposure, command injection where input becomes a command rather than a value, Lua or scripting evaluation on untrusted input, and dangerous administrative commands that rewrite the on-disk file or load a module. Use when an application talks to a cache or key-value store and either the instance is network-reachable or untrusted input reaches the command layer. The untrusted input or the open port is the source, the datastore command interface is the sink, and the unauthenticated access or the composed dangerous command is the bug.
- auditing-declarative-authorization
Audit authorization expressed as configuration or framework convention rather than inline code: row-level security and policy rules, framework before-action and middleware filters that must be attached to every protected route, serverless and gateway access rules, and object-ownership checks. Covers routes that skip the filter, policies with a permissive default, rules that check authentication but not ownership, and gaps between where the rule is declared and where the data is accessed. Use when reviewing role- or policy-driven access control. Coverage and correctness are separate checks.
- auditing-declared-vs-used-permissions
Find the consent gap in an agent skill or MCP server: the distance between the permissions and capabilities it declares and what its bundled code and instructions actually exercise. Covers over-broad grants a skill requests but never uses, capabilities it exercises without declaring, and grants that are used but still wider than the task needs. Read the declared surface in frontmatter or manifest, inventory the real behavior, and diff the two in both directions. Use when reviewing a skill or server before install, or auditing least privilege across an agent's installed set. An over-broad or undeclared grant is the finding.
- auditing-device-code-and-pkce-flows
Audit the server side of the authorization-code-with-proof-key and device-authorization grants for bugs that let a stolen or guessed code become a token. Covers a token endpoint that issues without checking the proof-key verifier at all, that accepts the plain challenge method or a challenge-absent downgrade, or that binds the verifier to the client rather than to the specific code; and a device grant whose short user code is brute-forceable because polling is unthrottled, whose device code is not bound to the requesting client, or whose approval is not tied to the authenticated approver. Scoped to the proof-key and device-code specifics, not general federated login, which a separate skill covers. Use when reviewing a token endpoint or a device-authorization endpoint. The token request parameters are the source, token issuance is the sink, and an unenforced proof binding between them is the bug.
- auditing-directory-sync-trust
Audit bulk directory synchronization (LDAP, HR-system, IdP, or cross-directory feeds) between an external identity source and an application for trust misplaced in the sync feed: a sync that trusts a source attribute (group, department, an admin-like flag) to set local privilege or tenancy without validating it, a connector authenticated by a broad credential that can read and reshape the whole directory, a mapping that lets an external group name land on a privileged internal group, a sync that matches accounts by a spoofable key (email, external id) so an attacker record merges into an existing identity, and a source deletion that does not propagate so departed users linger. Use when an external directory feed drives account and privilege state and the application's trust in that feed is the boundary. The attacker-influenced source record is the source, the over-privileged, merged, or lingering internal account is the sink, and the unvalidated attribute mapping or spoofable match key is the bug.
- auditing-ecs-task-metadata-boundaries
Audit container task credential and metadata boundaries in orchestrated compute such as ECS: a workload that can reach the container credential endpoint or the host instance metadata service to obtain a role broader than the task needs, a task role over-scoped for the workload, a sidecar or co-located container sharing the same credentials, and a server-side request path inside the task that reaches the metadata endpoint. Covers the task credential relative URI, the instance metadata service reachable from a task, and the blast radius when one container in a task is compromised. Use when containerized workloads assume a task or instance role and the metadata endpoints are the boundary. The reachable metadata endpoint is the source, the credential it returns is the sink, and the role wider than the task's need is the bug.
- auditing-editor-extension-workspace-trust
Audit an editor or IDE extension for actions it runs on untrusted workspace contents, after the workspace-trust capability and activation events are resolved. Covers a task, debug preLaunchTask, or command that auto-runs on folder open, a tool path or command template read from workspace settings and executed, a workspace-trust bypass where a risky operation runs in restricted mode, a language server that loads or executes a workspace-declared binary, repo-controlled data reaching a shell or task-execution sink, and an extension webview that renders repo content without a content-security policy. Use when reviewing an extension manifest (activation events, the untrusted-workspace capability), its workspace config handling, or its language-server tool resolution. A cloned untrusted repository is the source, an action the extension auto-runs on open is the sink, and repo-controlled code reaching execution without a trust gate is the bug.
- auditing-electron-ipc-trust
Audit an Electron desktop app for untrusted renderer content that reaches a Node or operating-system capability, after the webPreferences and the preload bridge surface are resolved. Covers nodeIntegration enabled with contextIsolation off, a preload that exposes raw ipcRenderer or a generic invoke passthrough, an ipcMain handler that trusts renderer input as a path, command, or URL, remote or attacker-influenced content loaded through loadURL with navigation unlocked, shell.openExternal called on a renderer-controlled string, and a custom protocol or deeplink routed into a privileged action without validation. Use when reviewing webPreferences, the preload and contextBridge surface, IPC handlers, and remote-content loading, not renderer-side markup injection the client-side DOM skill owns. Untrusted content in a renderer is the source, a Node or operating-system capability is the sink, and input crossing the bridge without validation while isolation is off is the bug.
- auditing-error-handling-and-information-exposure
Audit error handling and diagnostic surfaces for sensitive information a real client receives, where an exception path, a debug feature, or a diagnostic endpoint returns stack traces, database errors, internal paths, framework or version banners, configuration, secrets, or messages that differ enough to enumerate users. Use when reviewing how a service responds to malformed, unauthorized, or failing requests in its deployed configuration, and whether debug modes, source maps, or version-control metadata are exposed. Scoped to what production actually returns, not developer-only verbosity. The error or diagnostic path is the source, the response, header, or user-visible log is the sink, and disclosing detail that aids a further attack is the bug.
- auditing-file-upload-and-content-handling
Audit a file-upload and content-handling path for an attacker-supplied file whose bytes, declared type, name, or embedded content reach a sink that stores it in a served or executable location or feeds it to a parser that acts on its content, after the type-decision layer and the serve behavior are resolved. Covers an extension or content-type trusted for a type decision that a second layer contradicts, an SVG, HTML, or XML file stored and served inline as active content, image or document parser exploitation, a polyglot file passing one content check yet executing in another context, an upload path writing outside the intended directory, and an archive expanding to a write primitive. Use when reviewing upload validation, storage, and serving, not the client-side DOM sink or the archive-extraction write primitive their own skills own. An uploaded file is the source, a serve or parse sink acting on its content is the sink, and a type decision one layer contradicts is the bug.
- auditing-graphql-attack-surface
Audit the attack surface a GraphQL API exposes that a plain endpoint does not: schema introspection left open, unbounded query depth and recursion, aliasing and field duplication that multiply cost, query batching that defeats rate limits and enables brute force, field-level authorization that a resolver skips even when the object check passed, and mutations reached without the guard their action needs. Covers the query and variables as the source, the resolver and the data or work it triggers as the sink, and the missing depth, cost, batch, or field guard as the bug. Use when reviewing a GraphQL schema, its resolvers, or a gateway that fronts one. Introspection and cost limits are one audit; per-field and per-mutation authorization is the other.
- auditing-group-policy-and-sysvol-trust
Audit trust placed in group policy content and the domain policy share, where a low-privileged principal can write a policy object, its files on the domain share, or its link, causing that policy to run scripts, set scheduled tasks, install software, or change security settings on every machine the policy applies to, or where a stored credential or an unsigned policy file lets any domain user act on the policy. Use when policy objects, their share files, or their links may be writable by non-administrators or carry stored secrets. Covers writable policy content, writable links and scope, embedded credentials, and unsigned or world-readable policy files. The low-privileged write to policy content or link is the source, the machines applying the policy are the sink, and running attacker-chosen policy on those machines is the bug.
- auditing-grpc-service-authorization
Audit a gRPC service for a method a caller can reach without the authorization the service assumes an interceptor enforces, after the interceptor coverage and the channel credentials are resolved. Covers authorization installed on the unary interceptor while the streaming chain omits it, a per-method authorization gap reachable at the wrong privilege, server reflection enabled in production exposing the full API, a plaintext channel with metadata trusted unverified, an absent message-size or recursion-depth limit inviting decode denial of service, and a transcoding gateway that does not apply the same auth filter as native gRPC. Use when reviewing service and interceptor registration, method handlers, and channel setup, not the certificate-validation mechanics the transport skill owns. A caller with forged or absent metadata is the source, a service method acting without an authorization check is the sink, and an interceptor that does not cover the method or stream is the bug.
- auditing-guard-gaps
Find the missing-check bug by comparing sibling functions that reach the same sink - one validates its input, its peer does not. Use on an authorized source target to surface broken access control, missing bounds checks, and skipped sanitization that linear file-reading hides; when you suspect one handler in a family forgot the check its siblings all perform. Covers finding a guarded anchor, enumerating structural peers, diffing guard-for-guard by what each actually enforces, and confirming the unguarded peer is reachable with attacker input.
- auditing-host-mount-and-device-exposure
Audit the host paths and devices a workload mounts for reach across the container boundary onto the node: a writable hostPath into a sensitive node directory, a mount of the host root or a system path that exposes other pods' data and node configuration, a raw device or block volume that grants low-level host access, and a hostPath whose subpath or symlink handling lets a workload escape the intended directory. Covers Kubernetes and container hosts where hostPath volumes and device mounts connect a container to the node filesystem and hardware. Use when workloads mount host paths or devices and the question is what node state that mount exposes. The workload holding the host mount is the source, the node file or device it reaches is the sink, and the host-path or device exposure beyond the workload's need is the bug.
- auditing-http2-and-grpc-multiplexing-trust
Audit HTTP/2 and gRPC edges for framing and multiplexing trust that breaks when a stream is translated or reused: an h2c or HTTP/2-to-HTTP/1.1 downgrade that reintroduces request smuggling, pseudo-header and header handling that lets a stream forge its path or authority, multiplexed streams on one connection whose authentication or rate limit is applied per connection rather than per stream, and a gRPC gateway that trusts metadata or a method name a caller controls. Covers HTTP/2 front ends, gRPC services, and gateways that translate between protocols. Use when an edge terminates or downgrades HTTP/2 or multiplexes gRPC calls and per-stream trust is assumed. The crafted stream or metadata is the source, the back-end request or method it reaches is the sink, and the downgrade or per-connection trust that admits it is the bug.
- auditing-iac-module-and-provider-supply-chain
Audit the supply chain of infrastructure-as-code modules and providers for trust that runs at plan or apply time: a module sourced from an unpinned or attacker-influenceable location, a provider or plugin pulled from a registry without integrity pinning, a module that executes local commands or fetches remote content during planning, and a lockfile that is missing, ignored, or not enforced in CI. Covers Terraform and similar declarative tools where a module or provider runs with the credentials of whoever applies it. Use when infrastructure is built from third-party or shared modules and providers and the apply identity is privileged. The untrusted module or provider source is the source, the plan or apply execution is the sink, and the code running under the applier's credentials without integrity pinning is the bug.
- auditing-idp-initiated-flow-trust
Audit identity-provider-initiated single sign-on for trust placed in an unsolicited assertion the application never asked for: an IdP-initiated SAML response the service provider accepts with no matching request so there is no request state to bind it to, an unsolicited assertion an attacker captures and replays or delivers to a victim to log them into an attacker-chosen account, a RelayState value trusted as a redirect target so it becomes an open redirect, an assertion with no or a too-wide audience so it is accepted by a service provider it was not meant for, and a missing replay defense (no one-time-use, weak expiry) that lets one assertion be used more than once. Use when an application accepts a login assertion it did not request and the validation of that unsolicited assertion is the boundary. The unsolicited or replayed assertion is the source, the unintended authenticated session is the sink, and the missing request binding, audience, or replay defense is the bug.
+154
Using security-agent-skills in Claude Code
Skills need no command. Claude reads them when your request matches what they describe.
Update, turn off or remove
- Update the plugin (shell)
- claude plugin update security-agent-skills@unboundcompute
- Refresh the marketplace
- /plugin marketplace update unboundcompute
- Turn it off for now
- /plugin disable security-agent-skills@unboundcompute
- Remove it
- /plugin uninstall security-agent-skills@unboundcompute
Skills in security-agent-skills
- adjudicating-dependency-cve-reachabilityUnboundCompute/security-agent-skills
Decide whether a CVE in a dependency actually exposes your application before you scramble to patch: is the vulnerable function on a real call path from your code, do the trigger preconditions hold, and can an attacker control the input that reaches it. Use when an advisory, SCA scan, or dependency bot flags a CVE and you must separate a genuine exposure from unreachable noise, or justify why you are or are not affected. Covers pinning the vulnerable symbol, call-path reachability, precondition checks, and taint from your entry points.
27stars - adjudicating-taint-pathsUnboundCompute/security-agent-skills
Decide whether a whitebox lead is a real bug by tracing taint from an untrusted source to a dangerous sink and confirming every hop against live source. Use after a scanner, a candidate list, or your own reading surfaces a "this looks dangerous" sink (SQL exec, system/exec, file open, deserialize, template render, redirect target, memcpy) and you must decide whether attacker- controlled input actually reaches it - or kill the lead with evidence. Covers forward and reverse taint, witness paths, sanitizer analysis, and the evidence rules that separate a finding from a false positive.
27stars - auditing-account-abstraction-and-paymaster-trustUnboundCompute/security-agent-skills
Audit an ERC-4337 account-abstraction deployment for trust misplaced in the user-operation lifecycle: a smart account whose validation accepts a signature or nonce it should reject, a paymaster that agrees to sponsor gas for operations it should not so an attacker drains its deposit, a bundler or entry-point assumption that lets a user operation be replayed or reordered for gain, and validation logic that reads mutable state or reaches outside its allowed scope. Covers smart-contract wallets, paymasters, bundlers, and the entry point in an account-abstraction stack where a user operation is validated and sponsored before it executes. Use when a user operation is validated, paid for, and executed by separate parties and that trust split is the boundary. The crafted user operation is the source, the drained paymaster or unauthorized execution is the sink, and the over-permissive validation or sponsorship rule is the bug.
27stars - auditing-account-recovery-and-reset-trustUnboundCompute/security-agent-skills
Audit password reset and account recovery flows for the trust that lets an attacker take over an account: a reset token that is guessable, long-lived, reusable, or not bound to the account it was issued for, a recovery path that verifies a weaker factor than login and bypasses multi-factor, a reset link whose host comes from an attacker-controllable header so the token leaks, and a recovery that trusts an unverified email or phone change to redirect the reset. Covers the recovery surface of authentication systems, where resetting a credential or recovering access is the alternate door into an account. Use when an application offers password reset or account recovery and that flow is a path to authentication. The attacker-driven recovery request is the source, the account it takes over is the sink, and the weak token, bypassed factor, or leaked reset link that grants it is the bug.
27stars - auditing-admission-control-policy-gapsUnboundCompute/security-agent-skills
Audit cluster admission control for gaps that let a non-compliant or hostile workload through: a validating webhook that fails open when its backend is unreachable, a policy that scopes by namespace or label and misses the namespaces that matter, a mutating webhook whose changes are trusted downstream, and an ordering or exemption that lets a privileged pod bypass the checks meant to stop it. Covers Kubernetes admission webhooks and policy engines that are supposed to enforce workload security at create and update time. Use when admission policies are the control that keeps privileged, unsigned, or over-permissioned workloads out of the cluster. The workload submitted for admission is the source, the admitted resource is the sink, and the policy gap that let it through is the bug.
27stars - auditing-ai-agent-permissionsUnboundCompute/security-agent-skills
Audit what an AI agent is actually allowed to do versus what its task needs. Covers excessive agency (tools, scopes, and autonomy beyond the job), missing human-in-the-loop gates on irreversible actions, over-broad credentials and their blast radius, sandbox and code-interpreter escape, unfiltered egress, and unbounded resource or spend (denial-of-wallet). Use when granting an agent a tool or scope, reviewing an agent's permission posture, or deciding which actions need approval. The model's restraint is not a control; permissions are.
27stars - auditing-android-component-exposureUnboundCompute/security-agent-skills
Audit an Android app for components another app on the device can reach and drive, after the manifest export flags and permission gates are resolved. Covers an activity, service, broadcast receiver, or content provider exported without a permission gate, an intent filter that makes a component implicitly exported, a provider that grants URI access or exposes files across the app boundary, a permission declared with a weak protection level, and a component that trusts intent extras from an untrusted caller for a sensitive action. Use when reviewing the manifest and the component code that handles an inbound intent, not the deep-link URL trust that a WebView loads (that is the deep-link skill). A cross-app caller is the source, a reachable exported component acting on the intent is the sink, and a sensitive component another app can invoke unguarded is the bug.
27stars - auditing-android-intent-redirection-and-pendingintentUnboundCompute/security-agent-skills
Audit Android privilege and access leaks through intent redirection and mutable pending intents, where a privileged component receives an intent carrying a nested intent it then launches, or hands out a pending intent an untrusted app can fill in, so the untrusted caller reaches a protected component or acts with the privileged app's identity, because the component forwards an attacker-supplied intent without constraint or the pending intent is mutable and under-specified. Use when an Android app forwards intents extracted from input or shares pending intents with other apps. Covers nested-intent redirection to protected components, mutable pending intents, implicit pending-intent delivery, and forwarding that carries the app's permissions. The attacker-supplied nested or fillable intent is the source, the privileged component launching or sending it is the sink, and reaching a protected target with the app's identity is the bug.
27stars - auditing-ansible-become-and-vault-trustUnboundCompute/security-agent-skills
Audit configuration-management privilege escalation and secret handling for trust that runs as root on every managed host: a task that escalates with become across a whole play when only one step needs it, a role or variable sourced from an untrusted place that runs under that escalation, a vault-encrypted secret whose decryption key is exposed to the runner or logged, and a templated value or module argument that takes attacker-influenceable input while privileged. Covers Ansible and similar agentless tools where a control node runs plays that escalate privilege and decrypt secrets across a fleet. Use when playbooks escalate with become or handle vault secrets across managed hosts. The untrusted role, variable, or input running under escalation is the source, the privileged task or decrypted secret is the sink, and the over-broad escalation or exposed key is the bug.
27stars - auditing-api-key-and-token-lifecycleUnboundCompute/security-agent-skills
Audit the lifecycle of API keys and access tokens for weaknesses that let one keep working past its intended bounds: a key issued with broader scope than the caller needs so a leak grants far more than one function, a key or token with no expiry that stays valid indefinitely, a revocation path that does not actually stop the key so a rotated or compromised credential keeps authenticating, a key that leaks into logs, URLs, client-side code, or error messages and is never rotated, and a token whose scope or audience is not enforced on use so it works against endpoints it was never meant for. Use when a long-lived programmatic credential authenticates a caller and the bounds on that credential (scope, expiry, revocability) are the boundary. The over-scoped, unexpiring, or leaked key is the source, the access it grants past its intended bounds is the sink, and the missing scope limit, expiry, or working revocation is the bug.
27stars - auditing-ble-and-gatt-authorizationUnboundCompute/security-agent-skills
Audit a Bluetooth Low Energy device for missing authorization on its GATT attributes: a characteristic performing a sensitive action or revealing sensitive data readable or writable by any peer, pairing or bonding that is not required or falls back to an unauthenticated Just Works mode with no link encryption, an authorization decision the device pushes to the mobile app instead of enforcing on the peripheral, and a replayable command a sniffer can capture and resend. Covers BLE peripherals, wearables, locks, medical and IoT devices, and their GATT services where a connected peer reads or writes characteristics. Use when a peer can connect over BLE and the peripheral's own enforcement of who may read or write each characteristic is the boundary. The unauthorized connected peer or captured command is the source, the sensitive read, write, or action on the peripheral is the sink, and the missing pairing, characteristic-level authorization, or replay protection is the bug.
27stars - auditing-break-glass-account-trustUnboundCompute/security-agent-skills
Audit emergency break-glass and privileged-access accounts for the ways their standing power outlives the emergency they exist for: a break-glass account with a static, shared, or never-rotated credential that is a permanent superuser login rather than a sealed last resort, an emergency account excluded from the MFA, conditional-access, or logging controls covering other admins so its use is easy and invisible, a just-in-time elevation that grants more than the task needs or is not revoked when the task ends, an emergency-access path with no approval, time bound, or alerting so it is used routinely, and a break-glass use that triggers no review afterward. Use when a standing emergency identity or an elevation path holds power beyond ordinary admins and the controls around when and how it is used are the boundary. The break-glass credential or elevation request is the source, the unbounded or unaudited privileged access is the sink, and the missing rotation, control coverage, or use-time alerting is the bug.
27stars - auditing-browser-extension-trustUnboundCompute/security-agent-skills
Audit a browser extension (Manifest V3) for a trust boundary another web page or extension can cross to reach a privileged capability, after the permission scope and the message-sender checks are resolved. Covers an externally_connectable or onMessageExternal handler that verifies an origin but not the calling script, a content-script-to-background message handler with no sender validation, host permissions broader than the extension needs, a web_accessible_resources page that acts on URL parameters, an injected content script writing page-controlled data to a DOM sink, and a weak or eval-permitting content-security policy. Use when reviewing the manifest, background and content scripts, and cross-context message passing, not the web-page DOM sink taxonomy the client-side DOM skill owns. An untrusted web origin or another extension is the source, a privileged extension API or DOM sink is the sink, and a message reaching it without a sender-and-origin check is the bug.
27stars - auditing-cicd-oidc-trustUnboundCompute/security-agent-skills
Audit continuous-integration pipelines for the trust they extend to untrusted input: workflows that run on incoming change requests from forks while holding repository secrets, steps that let attacker-controlled content reach a privileged command, and cloud role trust conditions that accept a pipeline's short-lived token too broadly. Covers secret and token exposure on fork-triggered runs, poisoned-pipeline execution, and over-broad trust on the identity claim a pipeline presents to a cloud account. Use when reviewing CI/CD configuration, pipeline identity, or the boundary between a build and the cloud it can reach. An exploitable token or command from untrusted input is the finding.
27stars - auditing-clickjacking-and-ui-redressingUnboundCompute/security-agent-skills
Audit a web application for UI-redressing attacks where an attacker frames the real site and tricks a user into acting on it unknowingly: a sensitive page that can be embedded in an attacker's iframe because it lacks frame-ancestors or X-Frame-Options, a state-changing action reachable by a single click that an overlay or transparent frame steers, a drag-and-drop or keystroke redressing that captures input meant for the attacker page, and a confirmation step that a framed overlay hides. Covers web pages with authenticated, state-changing actions (settings changes, purchases, approvals, connect flows) that could be loaded inside a frame the user cannot see. Use when a sensitive action can be triggered by a click and the page can be framed by another origin, making framing the boundary. The attacker page that frames or overlays the real site is the source, the unknowing state-changing click is the sink, and the missing framing protection or unguarded one-click action is the bug.
27stars - auditing-container-image-build-hardeningUnboundCompute/security-agent-skills
Audit container image build definitions (Dockerfile, containerfile, and the compose or run config that sets runtime flags) for an image that ships over-privileged or carrying a secret, after multi-stage discards and deploy-time overrides are accounted for. Covers an image that runs as root, a secret baked into a layer, remote content pulled unpinned or unverified, a mutable or untagged base, an over-broad copy that pulls in local secrets and history, and a dangerous runtime request such as privileged mode or a sensitive host mount. Use when reviewing the image build plane, not the deploy-time security context or the cloud resource definition. The build definition is the source, the shipped image or run config is the sink, and privilege or a secret that survives into the final image is the bug.
27stars - auditing-container-image-provenanceUnboundCompute/security-agent-skills
Audit how a cluster decides which container images to trust and run: an image referenced by a mutable tag rather than a content digest, a workload pulling from a registry that admits unsigned or unverified images, a signature or attestation policy that is configured but not enforced at admission, and a base image or layer whose origin the pipeline never verified. Covers Kubernetes and container platforms where the image a workload runs is the code that runs, and where tag mutability, signing, and provenance decide whether it is the intended artifact. Use when workloads pull images whose signing and provenance are not enforced end to end. The unverified image reference is the source, the running container is the sink, and the code that runs without proven provenance is the bug.
27stars - auditing-container-runtime-and-socket-exposureUnboundCompute/security-agent-skills
Audit whether a workload can reach the container runtime and thereby control the host: the container runtime socket mounted into a pod or bound into a build or CI container, a privileged sidecar that talks to the runtime to launch or inspect containers, a runtime API exposed over a reachable port, and tooling that needs the runtime and is given it far more broadly than the one operation requires. Covers container hosts where reaching the runtime socket or API grants the ability to start privileged containers, mount the host filesystem, and take over the node. Use when a workload, build, or agent is given access to the container runtime. The workload with runtime access is the source, the container runtime is the sink, and the socket or API exposure that grants host control is the bug.
27stars - auditing-cors-and-cross-origin-trustUnboundCompute/security-agent-skills
Audit the code and configuration that decide cross-origin access, for trust a browser turns into a read of authenticated data. Covers a response that reflects an arbitrary request origin into the allow-origin header alongside allow-credentials, an allowlist that accepts the null origin, allowlist matching by prefix, suffix, substring, or an unanchored pattern that a lookalike origin satisfies, the origin header trusted as an authorization or request-forgery defense, and a cross-window message handler that acts on data without an exact origin and source check. Scoped to the code and config that build the decision, not a live-header scan. Use when reviewing cross-origin response headers, origin-based access logic, or cross-window message handlers. The request origin or the posted message is the source, the credentialed cross-origin read or the message sink is the sink, and trusting an attacker-set origin is the bug.
27stars - auditing-cross-account-role-trust-boundariesUnboundCompute/security-agent-skills
Audit cross-account IAM role assumption for trust policies that let the wrong principal assume a role: a trust policy with a wildcard or overbroad principal, a missing or unverifiable external ID on a third-party role, a confused-deputy path where a vendor assumes your role on any customer's behalf, and a role chain that reaches privileges the origin principal should never hold. Covers AWS assume-role trust policies, condition keys that should scope who may assume, and the transitive reach of one assumption into the next. Use when roles in one account can be assumed from another account, a partner, or a service, and the trust policy is the boundary. The external principal permitted by the trust policy is the source, the assume-role grant is the sink, and the trust scope wider than the intended relationship is the bug.
27stars - auditing-cross-chain-bridge-and-message-trustUnboundCompute/security-agent-skills
Audit a cross-chain bridge or messaging protocol for misplaced trust in messages that cross chains: a destination contract that accepts a mint or release on a forged or unverified proof of a source-chain event, a message whose signer set or validator quorum can be spoofed or is too small, a message that can be replayed on the destination or across chains for a repeated withdrawal, and a lock-and-mint or burn-and-release accounting that a crafted message pushes out of balance. Covers token bridges, message-passing layers, and any protocol where an action on one chain is authorized by an event claimed to have happened on another. Use when a destination-chain action depends on trusting a source-chain event and that verification is the boundary. The forged or replayed cross-chain message is the source, the unauthorized mint, release, or state change is the sink, and the missing or spoofable source-event verification is the bug.
27stars - auditing-csrf-and-unsafe-state-changesUnboundCompute/security-agent-skills
Audit state-changing endpoints for cross-site request forgery, where a request that rides the victim's ambient cookies is authorized on that session alone with no unpredictable, session-bound proof the request came from the app, letting an attacker page trigger the change as the victim. Use when reviewing forms and actions that modify data, change settings, move funds, or alter access, and how each is protected. Covers missing or unvalidated tokens, tokens not bound to the session, cookie-only same-site reliance and its gaps, state-changing GET requests, and content-type or method-override assumptions. The cross-site request on the victim's session is the source, the state-changing endpoint is the sink, and acting without unpredictable session-bound proof of origin is the bug.
27stars - auditing-datastore-exposure-and-abuseUnboundCompute/security-agent-skills
Audit in-memory and cache datastores such as Redis and memcached for exposure and command abuse: an instance reachable without authentication, a request that composes datastore commands from untrusted input, or server-side scripting and module or config commands that reach code execution or a file write. Covers unauthenticated network exposure, command injection where input becomes a command rather than a value, Lua or scripting evaluation on untrusted input, and dangerous administrative commands that rewrite the on-disk file or load a module. Use when an application talks to a cache or key-value store and either the instance is network-reachable or untrusted input reaches the command layer. The untrusted input or the open port is the source, the datastore command interface is the sink, and the unauthenticated access or the composed dangerous command is the bug.
27stars - auditing-declarative-authorizationUnboundCompute/security-agent-skills
Audit authorization expressed as configuration or framework convention rather than inline code: row-level security and policy rules, framework before-action and middleware filters that must be attached to every protected route, serverless and gateway access rules, and object-ownership checks. Covers routes that skip the filter, policies with a permissive default, rules that check authentication but not ownership, and gaps between where the rule is declared and where the data is accessed. Use when reviewing role- or policy-driven access control. Coverage and correctness are separate checks.
27stars - auditing-declared-vs-used-permissionsUnboundCompute/security-agent-skills
Find the consent gap in an agent skill or MCP server: the distance between the permissions and capabilities it declares and what its bundled code and instructions actually exercise. Covers over-broad grants a skill requests but never uses, capabilities it exercises without declaring, and grants that are used but still wider than the task needs. Read the declared surface in frontmatter or manifest, inventory the real behavior, and diff the two in both directions. Use when reviewing a skill or server before install, or auditing least privilege across an agent's installed set. An over-broad or undeclared grant is the finding.
27stars - auditing-device-code-and-pkce-flowsUnboundCompute/security-agent-skills
Audit the server side of the authorization-code-with-proof-key and device-authorization grants for bugs that let a stolen or guessed code become a token. Covers a token endpoint that issues without checking the proof-key verifier at all, that accepts the plain challenge method or a challenge-absent downgrade, or that binds the verifier to the client rather than to the specific code; and a device grant whose short user code is brute-forceable because polling is unthrottled, whose device code is not bound to the requesting client, or whose approval is not tied to the authenticated approver. Scoped to the proof-key and device-code specifics, not general federated login, which a separate skill covers. Use when reviewing a token endpoint or a device-authorization endpoint. The token request parameters are the source, token issuance is the sink, and an unenforced proof binding between them is the bug.
27stars - auditing-directory-sync-trustUnboundCompute/security-agent-skills
Audit bulk directory synchronization (LDAP, HR-system, IdP, or cross-directory feeds) between an external identity source and an application for trust misplaced in the sync feed: a sync that trusts a source attribute (group, department, an admin-like flag) to set local privilege or tenancy without validating it, a connector authenticated by a broad credential that can read and reshape the whole directory, a mapping that lets an external group name land on a privileged internal group, a sync that matches accounts by a spoofable key (email, external id) so an attacker record merges into an existing identity, and a source deletion that does not propagate so departed users linger. Use when an external directory feed drives account and privilege state and the application's trust in that feed is the boundary. The attacker-influenced source record is the source, the over-privileged, merged, or lingering internal account is the sink, and the unvalidated attribute mapping or spoofable match key is the bug.
27stars - auditing-ecs-task-metadata-boundariesUnboundCompute/security-agent-skills
Audit container task credential and metadata boundaries in orchestrated compute such as ECS: a workload that can reach the container credential endpoint or the host instance metadata service to obtain a role broader than the task needs, a task role over-scoped for the workload, a sidecar or co-located container sharing the same credentials, and a server-side request path inside the task that reaches the metadata endpoint. Covers the task credential relative URI, the instance metadata service reachable from a task, and the blast radius when one container in a task is compromised. Use when containerized workloads assume a task or instance role and the metadata endpoints are the boundary. The reachable metadata endpoint is the source, the credential it returns is the sink, and the role wider than the task's need is the bug.
27stars - auditing-editor-extension-workspace-trustUnboundCompute/security-agent-skills
Audit an editor or IDE extension for actions it runs on untrusted workspace contents, after the workspace-trust capability and activation events are resolved. Covers a task, debug preLaunchTask, or command that auto-runs on folder open, a tool path or command template read from workspace settings and executed, a workspace-trust bypass where a risky operation runs in restricted mode, a language server that loads or executes a workspace-declared binary, repo-controlled data reaching a shell or task-execution sink, and an extension webview that renders repo content without a content-security policy. Use when reviewing an extension manifest (activation events, the untrusted-workspace capability), its workspace config handling, or its language-server tool resolution. A cloned untrusted repository is the source, an action the extension auto-runs on open is the sink, and repo-controlled code reaching execution without a trust gate is the bug.
27stars - auditing-electron-ipc-trustUnboundCompute/security-agent-skills
Audit an Electron desktop app for untrusted renderer content that reaches a Node or operating-system capability, after the webPreferences and the preload bridge surface are resolved. Covers nodeIntegration enabled with contextIsolation off, a preload that exposes raw ipcRenderer or a generic invoke passthrough, an ipcMain handler that trusts renderer input as a path, command, or URL, remote or attacker-influenced content loaded through loadURL with navigation unlocked, shell.openExternal called on a renderer-controlled string, and a custom protocol or deeplink routed into a privileged action without validation. Use when reviewing webPreferences, the preload and contextBridge surface, IPC handlers, and remote-content loading, not renderer-side markup injection the client-side DOM skill owns. Untrusted content in a renderer is the source, a Node or operating-system capability is the sink, and input crossing the bridge without validation while isolation is off is the bug.
27stars
security-agent-skills plugin questions
How do I install the security-agent-skills plugin?
In Claude Code run /plugin marketplace add UnboundCompute/security-agent-skills, then /plugin install security-agent-skills@unboundcompute.
What does the security-agent-skills plugin include?
security-agent-skills bundles 194 skills.
How do I update or remove the security-agent-skills plugin?
Update it with claude plugin update security-agent-skills@unboundcompute in your shell, or from the Installed tab of /plugin. Remove it with /plugin uninstall security-agent-skills@unboundcompute.
What license is the security-agent-skills plugin under?
MIT, as stated in the plugin's manifest.