docs-init skill
✓ VerifiedBootstraps a documentation structure by exploring the codebase and populating templates with verified information.
- GitHub stars
- 8
- Files
- 3
- Last commit
- Oct 28, 2026
- License
- MIT
- Allowed tools
- —
What does the docs-init skill do?
docs-init bootstraps a documentation structure by exploring the codebase and populating templates with verified information.
Summary generated from the skill's SKILL.md.
About the docs-init skill
Setting up a comprehensive documentation folder for an undocumented codebase is what docs-init enables. It is designed for developers starting new projects or managing existing repositories that lack a docs/ directory. The tool explores the codebase by searching for package manifests, database models, and infrastructure configurations to determine the project's specific characteristics.
Once the technology stack is identified, the skill populates templates from assets/templates/ to create files such as architecture.md, onboarding.md, and data-model.md. It follows strict anti-hallucination rules, ensuring every claim is backed by actual code findings rather than assumptions. It also attempts to validate any Mermaid diagrams using the mmdc tool to ensure they render correctly. This process is intended for initial scaffolding rather than updating existing documentation.
What an agent can do with it
- 1Scaffold new project documentationInitialize a complete documentation structure for a brand new repository from scratch.
- 2Automate deployment and security docsProduce deployment and security documentation when infrastructure files like Dockerfiles are present.
- 3Verify diagram syntaxEnsure that any Mermaid diagrams included in the new documentation are syntactically correct.
Written from the skill's SKILL.md. Every use case is checked against a quote from it.
How to install the docs-init skill
Each command copies the skill folder into the client's skills directory. Restart the client or start a new session to load it.Install the docs-init skill in Plugin
/plugin marketplace add mgiovani/cc-arsenal /plugin install cc-arsenal@cc-arsenal-marketplace
Install the docs-init skill in Claude Code
~/.claude/skills/docs-init/SKILL.md
$ npx degit mgiovani/cc-arsenal/skills/docs-init ~/.claude/skills/docs-initInstall the docs-init skill in Claude Code (project)
.claude/skills/docs-init/SKILL.md
$ npx degit mgiovani/cc-arsenal/skills/docs-init .claude/skills/docs-initInstall the docs-init skill in Codex
~/.agents/skills/docs-init/SKILL.md
$ npx degit mgiovani/cc-arsenal/skills/docs-init ~/.agents/skills/docs-initInstall the docs-init skill in Gemini CLI
~/.gemini/skills/docs-init/SKILL.md
$ npx degit mgiovani/cc-arsenal/skills/docs-init ~/.gemini/skills/docs-initInstall the docs-init skill in Claude.ai
npx degit mgiovani/cc-arsenal/skills/docs-init docs-init && zip -r docs-init.zip docs-init # claude.ai → Customize → Skills → + → Create skill → Upload a skill → docs-init.zip
Automated checks
Ran Nov 15, 2026Full report →Passed our automated checks. That is not a security audit: read the scripts before you let an agent run them. How checks work · What to fix for more points
- Source repository linked and reachablemgiovani/cc-arsenal/skills/docs-init is public and reachablegithub.com/mgiovani/cc-arsenal/tree/HEAD/skills/docs-init ↗Pass
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit 2026-10-28github.com/mgiovani/cc-arsenal/commits ↗Pass
- No download-and-run or credential theft in scriptsNo scripts in the skill folderPass
- No hidden instructions in SKILL.mdNo override phrases, safeguard switches, credential requests or hidden charactersgithub.com/mgiovani/cc-arsenal/blob/HEAD/skills/docs-init/SK ↗Pass
- Pre-approved tools are scopedNo pre-approved tools; the agent asks before using onegithub.com/mgiovani/cc-arsenal/blob/HEAD/skills/docs-init/SK ↗Pass
- SKILL.md follows the Agent Skills formatthe description does not say when to use the skillgithub.com/mgiovani/cc-arsenal/blob/HEAD/skills/docs-init/SK ↗Finding
SKILL.md
View on GitHubInitialize Project Documentation
Bootstrap a docs/ structure for a project with little or no existing documentation. Only generate what the codebase actually evidences.
Anti-hallucination guidelines
Every claim in generated docs must trace to something read or grepped in this run:
- Verify a file/directory exists before referencing it.
- Get counts from
ls/find/grep, never estimate. - Quote real function/class/table names, not assumed ones.
- An empty directory is not a feature, don't document it.
- If a claim can't be verified, drop it rather than guess.
Workflow
1. Explore the codebase
Find: source directories with actual code, package manager files (package.json, pyproject.toml, go.mod, Cargo.toml, ...), database/ORM files, infrastructure configs (Docker, k8s, Terraform), and any existing docs/. See Detection Commands below for the exact patterns.
If a Task/subagent tool is available, delegate this to an Explore agent with that scope. Otherwise run the same grep/find commands inline and Read each hit: the result must be the same either way.
Verify every finding before using it: read the package manifest, read the model file, confirm a directory has real files inside it, not just an empty folder.
2. Detect project characteristics
From the exploration, determine: language/framework stack, project type (web app, CLI, library, service), whether a database/ORM is present, and whether infrastructure/deployment configs exist. These characteristics drive which doc types Phase 3 decides to generate.
3. Decide which docs to generate
- Always:
docs/architecture.md,docs/onboarding.md,docs/adr/0001-record-architecture-decisions.md(Nygard-format meta-ADR, generate inline, no template file) - If a database/ORM was found:
docs/data-model.md - If deployment configs were found (Dockerfile, k8s manifest, CI workflow, IaC):
docs/deployment.md,docs/security.md - If the project looks collaborative (multiple contributors in
git log, an open-source license, no existing CONTRIBUTING):docs/contributing.md,docs/rfc/directory
Don't generate a doc type with no supporting evidence: an empty data-model.md for a stateless CLI is worse than no file at all.
4. Check for existing docs
Scan docs/. For any target file that already exists, do not overwrite it: list it under "skipped" and ask the user before touching it. This skill is safe to rerun: by default it only fills gaps.
5. Populate templates
Templates live in assets/templates/ (see reference table below). For each one you're using:
- Grep it for its actual placeholder set:
grep -oE '\{\{[A-Z_0-9]+\}\}' assets/templates/<name>.md | sort -u - Map every placeholder to a value from step 1/2's verified findings. Never leave a placeholder as a literal
{{TOKEN}}in the output. - Templates mark some sections as conditional with an HTML comment ("delete if...", "only if evidenced"). Where the codebase gives no evidence for that section, delete the whole section (heading included, not just the placeholder text).
- After writing the file, grep it for
\{\{[A-Z_0-9]+\}\}again. Zero matches. If any remain, resolve or delete them before moving on.
Worked example (a FastAPI + PostgreSQL service):
$ grep -oE '\{\{[A-Z_0-9]+\}\}' assets/templates/architecture.md | sort -u
{{COMPONENT_DEPENDENCIES}}
{{COMPONENT_DESCRIPTION}}
{{COMPONENT_NAME}}
{{DATA_FLOW}}
{{DATE}}
{{DEPLOYMENT_SUMMARY}}
...
Map each to a verified finding: {{TECHNOLOGY_STACK}} becomes "Python 3.12, FastAPI 0.115, PostgreSQL 16 via SQLAlchemy", read from pyproject.toml and the model files, not assumed. If security.md's Compliance section has no GDPR/HIPAA evidence in the codebase (no consent flow, no PHI handling), delete that whole section rather than fill it with a guess.
6. Write the files
Create docs/ (and docs/adr/, docs/rfc/ if needed). Write each file that isn't being skipped.
7. Validate the diagrams render
Skip if none of the files you wrote has a ```mermaid block. Otherwise you need mmdc: install with npm i -g @mermaid-js/mermaid-cli, or run it ad hoc by defining mmdc() { npx -y -p @mermaid-js/mermaid-cli mmdc "$@"; } first. If Chromium refuses to launch in a sandbox, add -p with a puppeteer config of {"args":["--no-sandbox"]}.
Run this on the files you wrote; each FAIL <name>-N.mmd names the doc and the Nth mermaid block in it:
d=$(mktemp -d)
for md in <the docs/ files you wrote>; do
awk -v out="$d/${md//\//__}" '/^```mermaid/{f=out "-" ++n ".mmd"; next} /^```/{f=""} f{print > f}' "$md"
done
for f in "$d"/*.mmd; do mmdc -q -i "$f" -o "${f%.mmd}.svg" || echo "FAIL $f"; done
Fix each failing block using its parser error, then re-run until nothing prints FAIL. Cap at 3 rounds, then report the blocks that still fail. If mmdc can't be installed (no Node, offline), say so in the report and fall back to a careful syntax review of each diagram.
8. Report
List what was created, what was skipped (already existed), diagram validation result (rendered, or still failing, or unvalidated), and next steps.
Template Reference
| Document | Template | Generated when |
|---|---|---|
| Architecture | architecture.md |
always |
| Onboarding | onboarding.md |
always |
| First ADR | inline (Nygard format) | always |
| Data Model | data-model.md |
database/ORM detected |
| Deployment | deployment.md |
deployment config detected |
| Security | security.md |
deployment config detected |
| Contributing | contributing.md |
collaborative project |
Detection Commands
Run these directly (or hand them to the Explore agent from step 1):
# Language/framework
find . -name "package.json" -o -name "pyproject.toml" -o -name "go.mod" -o -name "Cargo.toml" | head -5
# Database/ORM
find . -name "*models.py" -o -name "*schema.prisma" -o -name "*entity.ts" | head -5
# Infrastructure
find . -name "Dockerfile" -o -name "docker-compose.yml" -o -name "*.k8s.yaml" | head -5
# Project name and description
basename "$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
head -20 README.md 2>/dev/null
Usage Examples
docs-init
docs-init for Python FastAPI microservice
docs-init for Next.js SaaS application
Example Output
Documentation Initialization Complete
Created:
docs/architecture.md - System architecture overview
docs/onboarding.md - Developer onboarding guide
docs/adr/0001-record-architecture-decisions.md - Meta-ADR
docs/data-model.md - Database schema (SQLAlchemy detected)
docs/deployment.md - Deployment guide (Docker detected)
Skipped (already exists):
docs/contributing.md
Next steps:
1. Review and customize the generated docs
2. docs-diagram er / docs-diagram arch for standalone diagrams
3. docs-adr "Decision Title" for future ADRs
Notes
- Safe to rerun: only fills gaps, asks before overwriting existing files.
- User-supplied context (e.g. "for a FastAPI microservice") steers detection but doesn't replace verification: still confirm the stack from the actual files.
Files in this skill
3docs-init skill questions
How do I install the docs-init skill in Claude Code?
Run npx degit mgiovani/cc-arsenal/skills/docs-init ~/.claude/skills/docs-init. Claude Code loads skills from ~/.claude/skills (all projects) or .claude/skills (one project) and uses a skill when its description matches the task.
Which clients can use the docs-init skill?
Any client that reads the Agent Skills format: Claude Code, Codex, Gemini CLI and claude.ai (upload as a ZIP). Scripts in the skill run only where the client allows code execution.
What happens if some documentation files already exist?
The skill will scan the docs/ directory and skip any target files that already exist, listing them under 'skipped' and asking the user before touching them.
Will the tool hallucinate information about my project?
No, the skill follows strict anti-hallucination guidelines, requiring every claim to be traced to something read or grepped in the codebase and deleting sections where no evidence is found.
Can this tool be used to update existing documentation?
No, this skill is specifically for initializing or scaffolding a new docs/ folder; you should use docs-update for refreshing or syncing existing docs.