Description
Matt Pocock Skills Reviewer success rate dropped from 100% (Aug 4-6, 9) to 54.5% (Aug 13, 11 runs, only 6 succeeded) per discussion #52498. The prompt .github/workflows/mattpocock-skills-reviewer.md hard-depends on invoking a pr-triage sub-agent and parsing its JSON response (Step 3) for skill selection, with no documented fallback if that call times out or returns invalid JSON - so one flaky sub-agent call can sink the entire run.
Expected Impact
A file-extension-based (or similarly simple heuristic) fallback for skill selection would prevent a single sub-agent dependency from causing full-run failures, directly addressing this cycle's regression.
Suggested Agent
Whoever owns .github/workflows/mattpocock-skills-reviewer.md.
Estimated Effort
Quick (< 1 hour) - add a default skill-selection branch when the pr-triage call fails/times out or returns unparseable JSON.
Data Source
DeepReport Intelligence Briefing 2026-08-13 analysis, discussion #52498 (Agent Performance Report - Week of 2026-08-13).
Generated by 🔬 Deep Report · agent · 253.2 AIC · ⌖ 48.7 AIC · ⊞ 11.4K · ◷
Description
Matt Pocock Skills Reviewer success rate dropped from 100% (Aug 4-6, 9) to 54.5% (Aug 13, 11 runs, only 6 succeeded) per discussion #52498. The prompt .github/workflows/mattpocock-skills-reviewer.md hard-depends on invoking a pr-triage sub-agent and parsing its JSON response (Step 3) for skill selection, with no documented fallback if that call times out or returns invalid JSON - so one flaky sub-agent call can sink the entire run.
Expected Impact
A file-extension-based (or similarly simple heuristic) fallback for skill selection would prevent a single sub-agent dependency from causing full-run failures, directly addressing this cycle's regression.
Suggested Agent
Whoever owns .github/workflows/mattpocock-skills-reviewer.md.
Estimated Effort
Quick (< 1 hour) - add a default skill-selection branch when the pr-triage call fails/times out or returns unparseable JSON.
Data Source
DeepReport Intelligence Briefing 2026-08-13 analysis, discussion #52498 (Agent Performance Report - Week of 2026-08-13).