A1 plan
Full planning pipeline for an implementation phase: Research → Map → Plan → Audit, producing an executor-ready PLAN.md with waves, tasks, and verifiable success criteria. State lives in `.a1/phases/<name>/`. MUST trigger when the user says: "plan phase" (alias: "phase planen"), "plan a new phase" (alias: "neue phase planen"), "create a plan" (alias: "plan erstellen"), "create PLAN.md" (alias: "PLAN.md erstellen"), "a1-plan", "plan this milestone", "plan M2-P1", "I need a plan for X", "let's plan the next phase", or any request to turn a goal/spec into an actionable, audited plan before execution. Auto-loops back through plan→audit on FAIL verdicts (max 2 cycles). Do NOT activate for: feature ideation from scratch (use a1-new-feature), bug fixes (use a1-fix), running an existing PLAN.md (use a1-execute), or codebase analysis without plan output (use a1-analyze).From its SKILL.md
npx -y skills add N3URAL-A1/a1-specforge --skill a1-planAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.2 KB, as published. Nobody here has run it
a1-plan — Phase Planning Pipeline
Language: English-first; German trigger aliases supported.
Thin orchestrator. Phase logic is in workflows/. Sub-agents do the thinking.
When to use
Activate when the user wants to plan a phase of implementation work — not execute it, not ideate it.
The output is a PLAN.md that's ready for a1-execute.
Phases
| # | Phase | Workflow | Agent | Output |
|---|---|---|---|---|
| 1 | Research | workflows/01-research.md | a1-rico-researcher | RESEARCH.md |
| 2 | Map | workflows/02-map.md | a1-marco-mapper | MAP.md |
| 3 | Plan | workflows/03-plan.md | a1-pablo-planner | PLAN.md |
| 4 | Audit | workflows/04-audit.md | a1-adam-auditor | AUDIT.md |
Audit loop: If AUDIT.md verdict is FAIL, route back to Phase 3 (a1-pablo-planner in revision mode) with the AUDIT.md findings. Maximum 2 revision cycles.
Retro: Phase 4 (Audit) closes with a mandatory Retro — see
workflows/04-audit.md — feeding a1-evolve's pattern clustering.
Storage
All artifacts live in .a1/phases/<phase-name>/ in the project directory:
.a1/
└── phases/
└── <phase-name>/
├── RESEARCH.md
├── MAP.md
├── PLAN.md
└── AUDIT.md
The phase directory is created automatically. Phase names are kebab-case from the user's description.
Routing
- Ask for: project path (default: current directory) + phase name/goal
- Create
.a1/phases/<name>/if not exists - Run Phase 1 → 2 → 3 → 4 in sequence (each phase needs the previous output)
- If AUDIT fails → Phase 3 (revision mode) → Phase 4 (re-audit)
- Present final PLAN.md location and brief summary
Hard rules
- Never skip phases — research and mapping take minutes and prevent rework
- Never edit PLAN.md directly — always go through a1-pablo-planner. Pablo owns
PHASE-level PLAN.md files under
.a1/phases/; FEATURE-level wave plans (vaultprojects/<slug>/plans/) are planned by a1-vincente-vibe-optimizer via a1-new-feature Phase 4. - If the user provides a spec file, pass its path to a1-rico-researcher and a1-pablo-planner
- Present the PLAN.md summary to the user after audit passes — they should confirm before executing