Request: Self-Review Skill and Ongoing Skill Governance

Request Summary

Add a user-invocable self-review skill where the agent evaluates improvements to skills and agent prompt quality, and keeps request tracking pages updated.

Scope

• Add user-visible /self-review skill.
• Require updates to Data/Pages/requests master list and grouped request pages.
• Keep recommendations categorized as now/next/later.

Status

Implemented

Evidence

• Added .github/skills/self-review/SKILL.md.
• Added Data/Pages/requests/skill-and-agent-prompt-requests-master-list-20260529.md.
• Added grouped request pages under Data/Pages/requests/.

Follow-Up

• Add cadence guidance (for example: weekly self-review pass).
• Link this request set to /tasks records for execution visibility.
• Treat chat tool usage parity as an explicit review dimension, not an implied prompt/runtime hygiene concern.
• Added from the 2026-05-29 follow-up self-review pass:
• grouped request page: skill-small-improvements-batch-20260529.md
• dedicated significant request page: pr-review-closure-and-thread-sync-skill-20260529.md
• dedicated significant request page: mcp-authoring-usability-and-chat-tool-parity-20260529.md
• task records: TSK-0223 and TSK-0224

2026-05-30 Follow-Up

Improvement shortlist (impact / effort)

1. Add a model-profile bootstrap hook that verifies Ollama tags and writes a trained-default + stock-comparator profile pair. - Impact: high - Effort: medium - Classification: Now - Confidence: 92%
2. Collapse duplicated council skill intent wording between repo and user-scoped skill surfaces into a single canonical contract note. - Impact: medium - Effort: low - Classification: Next - Confidence: 88%
3. Add a compact parity matrix output mode for prompt/runtime/tool-catalog checks to reduce token cost in recurring audits. - Impact: medium - Effort: low-medium - Classification: Next - Confidence: 89%
4. Add a monthly script-driven self-review cadence entrypoint that snapshots open request pages and unresolved task links. - Impact: medium - Effort: low - Classification: Later - Confidence: 84%

Evidence

• Local Ollama currently has stock qwen3.5:4b; trained memorysmith-athena:latest is not present in the current local runtime.
• Added a 24k stock comparator tag qwen3.5:4b-24k-stock and profile wiring in local override files to support A/B routing comparisons in Admin Models and chat profile selection.

Tracking

• Dedicated significant request page: model-profile-and-ollama-comparator-parity-20260530.md
• Significant execution task: TSK-0253