agentic-os/brain/skill-usage.md

2.1 KiB
Raw Blame History

Skill Usage & Effectiveness

Last updated: 2026-08-02

Source: audit/audit.log (373 entries) — 46 skill_run, 24 skill_scored (counted separately).

Most Used Skills (by audit log skill_run count)

Skill Runs Last Run Avg Score
memory-consolidation 13 2026-07-30 50 (n=2)
backup-skill 6 2026-08-02 48 (n=5)
code-review 5 2026-07-26 50 (n=3)
audit-test-plugin 5 2026-08-02 45 (n=5) — all runs errored
daily-standup 3 2026-07-31 45 (n=3)
tdd-cycle 2 2026-06-26
test-plugin 2 2026-06-28
goal-planner 2 2026-06-28
firebase-hosting-basics 2 2026-06-28
firebase-app-hosting-basics 1 2026-08-02 45 (n=1)
firebase-ai-logic-basics 1 2026-07-25 45 (n=1)
heartbeat 1 2026-07-25 50 (n=1)
brainstorming 1 2026-07-28 50 (n=1)

Most Effective Skills (by eval score, /100 scale)

Skill Avg Score Scored Runs Trend
devops-audit 53 1 flat
code-review 50 3 flat
memory-consolidation 50 2 flat
heartbeat 50 1 flat
brainstorming 50 1 flat
content-draft 50 1 flat
backup-skill 48 5 flat
daily-standup 45 3 flat
audit-test-plugin 45 5 flat — scores unreliable, all runs failed

⚠️ Scores are compressed in a 4553 band. The eval layer records failed opencode/Hermes-routed runs as success: true, so scores are not a trustworthy quality signal until routing is fixed.

Skills Per Project

Project Skills Used
Agentic OS memory-consolidation, backup-skill, daily-standup, heartbeat, devops-audit, audit-test-plugin, goal-planner, brainstorming, multi-angle-orchestration
Kitchen-Inventory code-review, systematic-debug, tdd-cycle, firebase-*
Pendelton-comms-live (none since 2026-02-28)

Learnings activity

10 skills carry learnings.md entries; no new learnings written since the 2026-07-30 memory-consolidation run. 9 skills have non-empty score-history.json.