known-good

name	known-good
description	Test fixture skill that satisfies all 15 auditor checks (Pass 1 + Pass 2). **Use when:** - The auditor integration tests assert PASS verdict on a canonical input **Triggers:** "known-good fixture", "auditor PASS smoke", "skill audit regression" **Not ideal for:** - Real-world use (this is a test fixture)
skill_api_version	1
context	{"window":"fork","intent":{"mode":"task"},"sections":{"exclude":["HISTORY"]},"intel_scope":"topic"}
metadata	{"tier":"meta","dependencies":[],"stability":"experimental"}
output_contract	stdout: known-good marker

/known-good — Auditor PASS test fixture

A minimal but complete SKILL.md that satisfies all 15 auditor checks. Used by tests/integration/test_skill_auditor.bats to confirm the auditor returns PASS or WARN on conforming input.

⚠️ Critical Constraints

• This is a test fixture, not a runtime skill. Do not invoke. Why: auditor tests pin the exact content; runtime use would surprise contributors.
• Frontmatter shape is canonical. YAML description: | block scalar with explicit Triggers and Use-when markers. Why: exercises description-has-triggers form (a) AND form (b) simultaneously.
• Constraints are front-loaded. This ## ⚠️ Critical Constraints section appears before any other H2. Why: tests constraints-frontloaded heuristic.
• Every constraint has a Why. Each bullet ends with **Why:** <reason>. Why: tests rationale-present heuristic.

Workflow

Phase 1: Setup

Read the fixture, parse frontmatter.

Checkpoint: frontmatter has name, description, metadata.tier.

Phase 2: Validate

Run all 8 Pass-2 checks.

Checkpoint: all 8 checks return pass or n/a.

Phase 3: Emit verdict

Aggregate via max-severity rule.

Checkpoint: verdict is PASS (no warns, no fails).

Output Specification

Format: stdout marker known-good confirming the fixture loaded cleanly. Filename: none — runtime fixture only. Exit code: 0.

Quality Rubric

• Frontmatter passes Pass 1 (heal structural)
• Constraints section is front-loaded (first H2 after title)
• Every constraint has a Why: rationale
• Each Phase has a Checkpoint
• Output Specification is explicit
• Quality Rubric exists with checkboxes
• SKILL.md is under 250 lines

Examples

# Run auditor against this fixture; expect exit 0 and verdict PASS or WARN
bash skills/skill-builder/scripts/audit.sh tests/fixtures/skills/known-good

Troubleshooting

Problem	Cause	Solution
Auditor returns FAIL	Fixture drifted from canonical shape	Restore from git; do not edit without updating tests