| name | quality-playbook |
| description | Repo-wide risk-proportional validation workflow for non-trivial changes with context, risk, validation, and follow-through. |
| category | validation |
| kind | policy |
| phase | verification |
| requires_write | false |
| requires_slurm | false |
| requires_benchmark_artifacts | false |
| delegates_to | [] |
| output_schema | skill_run_summary.v1 |
| aliases | ["quality-strategy","proof-policy"] |
Quality Playbook
Purpose
Provide a minimal proof-first workflow for higher-risk changes that need stronger validation and
explicit risk accounting without displacing research progress.
Workflow
- Restate objective, in-scope boundaries, and affected contracts.
- Classify risk:
- behavior, benchmark, docs, reproducibility, AI-workflow.
- Pick the smallest proving approach:
- targeted tests, canonical scripts, and representative sample runs.
- Make the minimal safe change set.
- Execute checks and document:
- what passed,
- what remains uncertain,
- any required follow-up issue.
Guardrails
- Avoid scope expansion and broad refactors during validation-sensitive tasks.
- Do not count fallback/degraded benchmark operation as success evidence.
- Keep follow-up work as explicit backlog items, not silent omissions.
Output
- Objective + risk class.
- Proof strategy and validation results.
- Residual risk and next required step.
When to use
Use this skill for the scope named in its frontmatter description and registry metadata.