master-skill

name	master-skill
description	Orchestrate end-to-end Agent Skill delivery as a standalone workflow with natural-language intent routing, optional /ss commands, session tracking, and deterministic verification.

Master Skill

Mission

Convert user intent into a complete standalone skill package with traceable lifecycle artifacts: IDEA -> DESIGN.md -> TASK.md -> IMPLEMENTATION -> VERIFY_REPORT.md -> ARCHIVE.

Mandatory Boot Sequence

1. Read knowledge/architect.md.
2. Read knowledge/standards.md.
3. Read loop/roadmap.md.
4. Resolve intent using natural language first, then optional /ss:* command hints.
5. Refuse ambiguous session context and ask at most three short clarifying questions.

Standalone Intent Router

• Natural language intent (primary):
  • Explore intent -> explore
  • Plan artifacts -> ff or continue
  • Execute tasks -> apply
  • Validate coherence -> verify
  • Finalize lifecycle -> archive
• Optional command triggers (secondary):
  • /ss:explore: Discover scope and constraints. Do not implement code.
  • /ss:new-session: Start a new standalone session context.
  • /ss:ff: Generate artifacts quickly until apply-ready.
  • /ss:continue: Create exactly one missing artifact per turn.
  • /ss:apply: Execute tasks and update checklist state immediately.
  • /ss:verify: Evaluate spec-task-code coherence.
  • /ss:archive: Close lifecycle after verification gates pass.
• Unsupported external command namespaces:
  • Reject unsupported namespace usage.
  • Return migration hint: Use /ss:* commands instead.

Ana-Skill Phase (Design)

1. Build or update DESIGN.md.
2. Capture strengths, weaknesses, information gaps, and risks.
3. Mirror design decisions into loop/roadmap.md.
4. Enforce Gate A and Gate B before moving forward.

Task-Skill Phase (Planning)

1. Build or update TASK.md.
2. Decompose work into phases, tasks, and checklist items.
3. Track task IO and Definition of Done per task group.
4. Mirror execution plan into loop/tasks.md.
5. Enforce Gate C before implementation.

Implementation Phase

1. Execute tasks in priority order.
2. Update - [ ] to - [x] as soon as a task is completed.
3. Keep changes inside required structure and conventions.
4. Stop and ask up to three short questions when blocked by missing context.

Moni-Skill Phase (Verification)

1. Build or update VERIFY_REPORT.md.
2. Evaluate Completeness, Correctness, and Coherence.
3. Classify issues by CRITICAL, WARNING, SUGGESTION.
4. Write pass/fail evidence into loop/checklist.md.
5. Enforce Gate D and Gate E before archive.

Phase Gates

• Gate A: Confirm scope and capabilities.
• Gate B: Freeze DESIGN.md.
• Gate C: Approve TASK.md with DoD.
• Gate D: Complete implementation tasks and validation.
• Gate E: Block archive while any CRITICAL issue exists.

Required Files

Maintain the following files at all times:

1. SKILL.md
2. knowledge/architect.md
3. knowledge/standards.md
4. loop/roadmap.md
5. loop/checklist.md
6. loop/tasks.md
7. scripts/init_skill.py
8. scripts/quick_validate.py
9. templates/DESIGN.md.template
10. templates/TASK.md.template
11. templates/VERIFY_REPORT.md.template

Response Contract

Return exactly eight sections in every execution response:

1. Mode
2. Using Session
3. Phase Focus
4. Action Taken
5. Files Created/Updated
6. Remaining Gaps
7. Risks / Blockers
8. Next Step

Operational Rules

1. Use imperative voice.
2. Apply progressive disclosure.
3. Trace every conclusion to requirement, spec, task, or code.
4. Deliver actionable outputs only.
5. Block archive whenever VERIFY_REPORT.md contains any CRITICAL issue.

Done Definition

Mark the skill complete only when all conditions pass:

1. Keep the structure compliant with required zones.
2. Describe Ana/Task/Moni workflow clearly in SKILL.md.
3. Ship required knowledge, loop, scripts, and templates.
4. Produce VERIFY_REPORT.md proving design-task-implementation coherence.
5. Obtain explicit user confirmation for final version closure.