Claude-skill-registry implement
Execute plan phases, writing code in spaces/ while tracking in ideas/. Use after creating a plan with /plan to implement work.
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/implement-taylorhuston-local-life-manager" ~/.claude/skills/majiayu000-claude-skill-registry-implement && rm -rf "$T"
manifest:
skills/data/implement-taylorhuston-local-life-manager/SKILL.mdsource content
/implement
Execute implementation phases from PLAN.md, writing code in
spaces/ while tracking progress in ideas/.
Usage
/implement yourbench YB-2 1.1 # Execute phase 1.1 of issue YB-2 /implement yourbench YB-2 --next # Auto-find next uncompleted phase /implement yourbench --next # Auto-detect issue + next phase /implement coordinatr 003 --full # Execute all remaining phases
The Bridge Pattern
ideas/yourbench/issues/YB-2-auth/ spaces/yourbench/ ├── TASK.md (requirements) ├── src/ ├── PLAN.md (phases) ←→ │ └── auth/ ← CODE WRITTEN HERE └── worklog/ └── tests/ ├── _state.json (current state) └── 001-phase-*.json (entries)
Prerequisites
REQUIRED:
must existissues/###-name/PLAN.md- Active issue context (issue number specified or inferable)
exists (code repository)spaces/{project}/
If PLAN.md missing: Run
/plan first to create implementation plan
Execution Flow
1. Parse & Validate
- Locate issue in
ideas/[project]/issues/###-*/ - Verify
existsspaces/[project]/ - Check git branch (warn if on main/develop)
- Check dependencies (warn if incomplete)
2. Load Worklog Context
Read
worklog/_state.json for:
- Current phase progress
- Key decisions made
- Previous agent context
- Blockers
3. Branch & Status Management
On first phase:
- Create feature branch:
orfeature/###-slugbugfix/###-slug - Update issue status to
in_progress - Initialize worklog directory
4. Execute Phase
- Load context:
- PLAN.md (phases and checkboxes)
- Spec section from
field in TASK.mdimplements: - ADRs and research docs
- Select agent based on phase domain
- Write code to
spaces/[project]/ - Run tests and quality gates
- Create worklog entry
- Update PLAN.md checkboxes
- Update
_state.json
5. Agent Coordination
| Phase Type | Primary Agent |
|---|---|
| Frontend UI | frontend-specialist |
| Backend API | backend-specialist |
| Database | database-specialist |
| Tests (RED) | test-engineer |
| Refactor | code-reviewer |
6. Per-Phase Security Checks (Conditional)
Trigger security-auditor after phase if touching:
- Authentication/Authorization
- Secrets/API keys
- Database operations
- File operations
- External APIs
- Deployment config
7. Worklog Entries (MANDATORY)
After each phase, create JSON entry:
- Filename:
{sequence:03d}-phase-{slug}.json - Required:
,type
,phase
,author
,summary
,work
,learnings
,next_stepsblockers
8. Final Review (--full mode)
When all phases complete:
- Launch code-reviewer agent
- Launch security-auditor agent
- Process combined results
- Block if CRITICAL issues
Modes
Specific Phase
/implement yourbench YB-2 1.1 # Execute only phase 1.1
Next Phase
/implement yourbench YB-2 --next # Find first uncompleted checkbox
Full Run
/implement yourbench YB-2 --full # All remaining phases + final review
Flags
| Flag | Purpose |
|---|---|
| Execute all remaining phases |
| Auto-detect next phase |
| Skip branch warnings |
| Skip per-phase security |
| Skip final reviews |
| Skip dependency warnings |
Workflow
/spec → /issue → /plan → /implement → /commit → /complete ↓ ↓ implements: Load spec section spec section for requirements ↓ worklog/ entries created ↓ spaces/[project]/ code written
Spec Compliance
When implementing, the code should fulfill the requirements from the spec section referenced in the issue's
implements: field. The /complete command will validate that all spec requirements are met.