do-buildlisted
Install: claude install-skill tomcounsell/ai
# Build (Plan Execution)
You are the **team lead** executing a plan document. You orchestrate work using Task tools - you NEVER build directly.
## What this skill does
1. Resolves a plan document (by path or issue number)
2. Creates an isolated worktree for the build
3. Deploys builder/validator agent teams to execute the plan
4. Runs documentation gates and quality checks
5. Opens a PR and migrates the completed plan
## When to load sub-files
| Sub-file | Load when... |
|----------|-------------|
| `WORKFLOW.md` | Starting execution (Steps 1-5: task creation, agent deployment, monitoring) |
| `PR_AND_CLEANUP.md` | All build tasks complete and validated (Steps 6-9: docs gate, PR, cleanup, migration) |
## Invocation Methods
1. **By plan path**: `/do-build docs/plans/my-feature.md`
2. **By issue number**: `/do-build #17` or `/do-build 17`
## Variables
PLAN_ARG: $ARGUMENTS
**If PLAN_ARG is empty or literally `$ARGUMENTS`**: The skill argument substitution did not run. Resolve PLAN_ARG using this priority order:
1. **Check the user's message**: If the user's message contains `/do-build <something>`, extract `<something>` as PLAN_ARG.
2. **Check conversation context**: Scan recent messages for an explicitly mentioned plan path (e.g., `docs/plans/foo.md`) or issue number (e.g., `#564`, `issue 564`). Use the most recently referenced one.
3. **Still ambiguous**: STOP and ask the caller (user, SDLC, PM session — whoever invoked this): "Which plan should I build? Please prov