forge skill
Implement tasks from a change. Use when the user wants to start or continue implementation of an existing change.
- GitHub 스타
- 13
- 파일
- 1
- 마지막 커밋
- 2026년 6월 23일
- 라이선스
- MIT
- 허용된 도구
- —
forge skill은 무엇을 하나요?
forge은(는) agent skill입니다. Implement tasks from a change. Use when the user wants to start or continue implementation of an existing change.
forge skill 설치 방법
각 명령은 skill 폴더를 클라이언트의 skills 디렉터리에 복사합니다. 클라이언트를 재시작하거나 새 세션을 시작하여 로드하세요.플러그인에 forge skill 설치
/plugin marketplace add devnomad-byte/techneering /plugin install tn@techneering
Claude Code에 forge skill 설치
~/.claude/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge ~/.claude/skills/forgeClaude Code (project)에 forge skill 설치
.claude/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge .claude/skills/forgeCodex에 forge skill 설치
~/.agents/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge ~/.agents/skills/forgeGemini CLI에 forge skill 설치
~/.gemini/skills/forge/SKILL.md
$ npx degit devnomad-byte/techneering/skills/forge ~/.gemini/skills/forgeClaude.ai에 forge skill 설치
npx degit devnomad-byte/techneering/skills/forge forge && zip -r forge.zip forge # claude.ai → Customize → Skills → + → Create skill → Upload a skill → forge.zip
SKILL.md
GitHub에서 보기Forge a Change
Implementation entry point. Load change artifacts, then chain to tn:blueprint for task breakdown and tn:assemble for execution.
Core Principle
READ ALL CONTEXT BEFORE WRITING ANY CODE.
Load every artifact. Understand the full picture. Then implement.
Steps
Step 1: Select the Change
If a name is provided, use it. Otherwise:
- Infer from conversation context if the user mentioned a change
- Auto-select if only one active change exists in
techneering/changes/ - If ambiguous, list all active changes with progress info and ask the user to select
Always announce: "Using change: "
Step 2: Load Context
Read all artifacts from the change directory:
techneering/changes/<name>/proposal.mdtechneering/changes/<name>/specs/**/*.mdtechneering/changes/<name>/design.mdtechneering/changes/<name>/tasks.md
Also read techneering/config.yaml for project context.
Step 3: Show Current Progress
Parse tasks.md for checkboxes: - [ ] (pending) vs - [x] (complete).
Display:
- Change name and schema
- Progress: "N/M tasks complete"
- Remaining tasks overview
Step 4: Create Implementation Plan (tn:blueprint)
Invoke the tn:blueprint skill to break down the tasks into bite-sized implementation steps:
- Exact file paths
- Complete code for each step
- Verification commands with expected output
- No placeholders
Step 5: Execute with Assemble (tn:assemble)
Invoke the tn:assemble skill to execute the plan:
- Each task gets a fresh subagent
- Two-stage review per task: spec compliance → code quality
- Independent tasks can run in parallel (max 3 concurrent)
- Follow
tn:redgreenfor each task (red-green-refactor) - Use
tn:diagnoseif issues arise
Step 6: Track Progress
After each task completes:
- Mark the corresponding checkbox in tasks.md:
- [ ]→- [x] - Report progress: "Task N/M complete"
Step 7: On Completion
Display:
## Implementation Complete
**Change:** <change-name>
**Progress:** N/N tasks complete
### Completed This Session
- [x] Task 1
- [x] Task 2
...
All tasks complete! Proceed to tn:gate to verify.
Step 8: On Pause (Issue Encountered)
Display:
## Implementation Paused
**Change:** <change-name>
**Progress:** N/M tasks complete
### Issue Encountered
<description>
**Options:**
1. <option 1>
2. <option 2>
What would you like to do?
Red Flags
| Thought | Reality |
|---|---|
| "I remember the requirements, no need to read files" | Lost context is the #1 cause of bugs |
| "Write code first, read spec later" | Spec is the source of truth — read it first |
| "This task is too simple for a blueprint" | All tasks go through blueprint → assemble |
| "Skip review to speed up" | Review is quality assurance — not optional |
Quick Reference
| Dimension | Key point |
|---|---|
| Role | Implementation entry point, chains blueprint → assemble |
| Input | Change name (optional, can auto-detect) |
| Internal chain | forge → blueprint → assemble (cannot be skipped) |
| Progress tracking | Mark tasks.md checkbox after each task |
| Path placeholder | Always use techneering/changes/<name>/ |
Next Step
forge is a composite skill that chains internally: tn:blueprint → tn:assemble, then exits to tn:gate. Users do not need to trigger the intermediate steps manually.
When done:
- All tasks complete → auto-invoke
tn:gate(then show menu) - Issue encountered → invoke
tn:diagnose - User wants to pause → save progress, wait for next session
- User requests stop → end the current flow
Guardrails
- Always read all context files before starting
- Keep code changes minimal and scoped to each task
- Update task checkbox immediately after completing each task
- Pause on errors, blockers, or unclear requirements — don't guess
- If implementation reveals design issues, pause and suggest artifact updates
- Unified path placeholder: always use
techneering/changes/<name>/
이 스킬의 파일
1forge skill 질문
Claude Code에 forge skill을 어떻게 설치하나요?
npx degit devnomad-byte/techneering/skills/forge ~/.claude/skills/forge를 실행합니다. Claude Code는 ~/.claude/skills(모든 프로젝트) 또는 .claude/skills(한 프로젝트)에서 skills를 로드하며, 설명이 작업과 일치할 때 skill을 사용합니다.
forge skill을 사용할 수 있는 클라이언트는 무엇인가요?
Agent Skills 형식을 읽는 모든 클라이언트(Claude Code, Codex, Gemini CLI, claude.ai(ZIP으로 업로드))가 사용할 수 있습니다. skill 내 스크립트는 클라이언트가 코드 실행을 허용하는 경우에만 실행됩니다.