By name: prd-vXX-skill-name description: > [1-2 sentence description of what this skill does]. Triggers on [specific phrases/contexts that should activate this skill]. Outputs [what the skill produces]. context: fork allowed-tools: - Read - Write - Edit - Glob - Grep
[Skill Name]
[One sentence summary of the skill's purpose.]
Workflow Overview
- Step 1 → Brief description
- Step 2 → Brief description
- Step 3 → Brief description
Core Output Template
[Define the primary output structure this skill produces]
| Element | Definition | Evidence |
|---|---|---|
| Field 1 | What this captures | How to validate |
| Field 2 | What this captures | How to validate |
See assets/output-template.md for copy-paste template.
Step 1: [First Major Step]
[Detailed instructions for this step]
Checklist
- Item 1
- Item 2
Step 2: [Second Major Step]
[Detailed instructions for this step]
Step 3: [Third Major Step]
[Detailed instructions for this step]
Quality Gates
Pass Checklist
- Gate 1: [Specific, measurable criterion]
- Gate 2: [Specific, measurable criterion]
- Gate 3: [Specific, measurable criterion]
Testability Check
- Can this output be validated?
- Does it reference IDs from SoT/?
How this is evaluated: This skill's output is scored by
scripts/readiness.py(a deterministic, $0, no-LLM proxy). Clearing a gate is permission to advance, not the goal — the goal is evidence. Padding inputs to move the score without moving the evidence is a frozen-replay defect. SeePRINCIPLES.mdP7.
Anti-Patterns
| Pattern | Example | Fix |
|---|---|---|
| Bad pattern 1 | "Example" | → "Better approach" |
| Bad pattern 2 | "Example" | → "Better approach" |
| Evidence fabrication | Back-filling confidence/evidence to clear a gate | → Cite a real source or mark the gap; never invent to pass (P4 anti-leakage, P7) |
Bundled Resources
references/[file].md— [Description of when to use]assets/[template].md— [Description of template purpose]
Handoff
[What happens after this skill completes? What stage/skill comes next?]