<< All versions

Skill v1.0.0

currentAutomated scan100/100
danielvm-git/bigpowers/e31-quality-guarantee
──Details
PublishedSeptember 30, 2026 at 02:18 PM
Content Hashsha256:0b64be4be12fe585...
Git SHA812d57a91767
──Files
Files (1 file, 2.8 KB)
SKILL.md2.8 KBactive
SKILL.md · 91 lines · 2.8 KB

version: "1.0.0"

STORY KEY: E31-S07 TITLE: Wire golden suite into evolve-skill regression checks TYPE: Story PARENT: e31 STATUS: Draft AUTHOR: dvm DATE: 2026-07-02 MATURITY: 3 SIZE: XS

1. Business narrative

evolve-skill currently uses run-benchmark for capability evaluations (measuring whether a skill improved). It lacks a fast regression check — if a skill edit breaks the sync pipeline or compliance, evolve-skill won't catch it until CI runs. Wiring run-golden-suite.sh as a pre-benchmark regression gate catches mechanical breakage before spending time on capability evals.

2. Value statement

As a skill author, I want evolve-skill to run the golden suite before benchmarking, so that mechanical regressions are caught before measuring capability changes.

3. Actors and permissions

  • Skill author (internal) — runs evolve-skill.
  • evolve-skill (system) — invokes golden suite as a gate.

4. Trigger and preconditions

Trigger: evolve-skill is invoked (benchmark-gated evolution flow). Precondition: scripts/run-golden-suite.sh exists (e31s03).

5. Main flow and business logic

  1. evolve-skill's SKILL.md references run-golden-suite.sh as a pre-benchmark step.
  2. Before running benchmarks, the agent executes run-golden-suite.sh.
  3. If golden suite fails, evolve-skill aborts before benchmark.

Interruption point: N/A.

6. Alternative flows and exceptions

6a. Golden suite not yet implemented — reference as "when available" with a note.

7. Interface elements

Context: existing (skills/evolve-skill/SKILL.md). Static elements: new process step referencing run-golden-suite.sh.

8. Domain model

Not applicable — documentation change only.

9. Integrations and boundaries

  • run-golden-suite.sh (perennial, direction: in) — invoked by evolve-skill.
  • run-benchmark (perennial, direction: out) — gated by golden suite.

10. Background processes

Not applicable.

11. Notifications

Not applicable.

12. Audit and logging

Not applicable.

13. Solution variabilities

Not applicable.

14. Quality attributes NFR

Not applicable — documentation only.

15. Security and compliance NFR

Not applicable.

16. UX and accessibility NFR

Not applicable.

17. Acceptance criteria

Scenario: evolve-skill references golden suite Given scripts/run-golden-suite.sh exists When skills/evolve-skill/SKILL.md is read Then it references run-golden-suite.sh for regression checks And it preserves the run-benchmark step for capability evals

18. Out of scope

  • Changing evolve-skill's benchmark flow.
  • Adding golden suite to other skills.

19. Open questions

Not applicable.

20. References

  • skills/evolve-skill/SKILL.md.
  • e31s03 (run-golden-suite.sh).
  • skills/run-benchmark/SKILL.md (capability evals).
All versions