[ScriptedAlchemy]

Problem

The usage recommender treats use_count <= patch_count as evidence for a repair hint after two patches. With unmeasured usage, the default zero satisfies that comparison, so merely patching a skill twice manufactures a repair recommendation. The stale-skill path already treats missing usage as unknown, but the improvement path did not.

Change

Require a recorded use count and timestamp before applying the existing observed-use patch-review heuristic. Missing or incomplete use evidence produces the existing non-actionable recommendation with an explicit evidence-gap reason, not a claim that no skill work is needed. Raw patch/view/use evidence remains available.

Pinned, user-authored, archived and disabled exclusions still run first. The observed-use policy, schemas, storage, MCP/dashboard response shape, writer execution, manual authoring and mutation authority are unchanged. The existing writer consumes this shared recommendation API, so missing telemetry no longer creates its managed_skill_patch hint.

Regression coverage

Updated the previous unmeasured-usage fixture to assert no repair hint while retaining all evidence. Added tests for incomplete count/timestamp pairs, the existing policy with observed use, and pinned/user-authored/disabled/archived exclusions with observed usage. The existing persisted-ledger integration test records a real Use event before its two patches and retains its positive expectation unchanged.

Validation

Reviewed the actual recommender, its writer consumer, MCP/dashboard call sites, existing persisted-ledger test, and the complete GitHub-generated commit diff. The diff changes only this recommender and its unit tests.

Rust compilation, test execution, rustfmt and Clippy have not run locally: this environment has no supplied checkout or Rust/Bazel toolchain. No Cargo commands were run. Hosted CI must execute the automation-runtime unit tests and the existing agent-suite skill-usage journey before merge. No model-quality or cost-saving measurement is claimed.

Refs #4538. This fixes the missing-telemetry inference only. Kev qualification/shadow evaluation, the separate keyword diagnosis, and deciding whether to retire the remaining observed-use heuristic remain open.

---

[changeset-bot]

⚠️ No Changeset found

Latest commit: 18c71bb

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

[ScriptedAlchemy]

Consolidated into #4742 (integrate/open-pull-requests) at the captured head shown in its description, as requested. Continue review and further fixes on #4742. This closure does not assert CI readiness; the replacement remains draft. Source branch retained.