agentleFS
Sign inStart free
wrg32786wrg32786/aigent-os/skills/agent-fitness/SKILL.md

agent-fitness

Skill26 starsChanged 10 days agoMITView on GitHub

Track sub-agent dispatch calibration from JSONL or ledgers. Use when checking agent fitness, clean/blocked/error trends, or which agents work.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    After major sessions where multiple sub-agents fired (today S30d had ~7 dispatches)

    line 40
  2. Distinctive:

    Before relying on a sub-agent for something heavy ("am I about to dispatch an agent that's been failing?")

    line 42
  3. Distinctive:

    Does not auto-reroute dispatches based on calibration.

    line 46
  4. Distinctive:

    Does not LLM-classify failures — heuristic only (keyword matching on result content).

    line 47
  5. Distinctive:

    Does not pool across sessions for cluster analysis.

    line 48
  6. Distinctive:

    Does not include task-type clustering.

    line 49

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/agent-fitness

Three modes:

/agent-fitness extract — runs python3 daemons/agent-fitness-extract.py to scan the latest JSONL transcript and append new dispatch rows to memory/AGENT_FITNESS.md. Idempotent (dedup by session_id + tool_use_id).

/agent-fitness (no args) — runs python3 daemons/agent-fitness-report.py. Surfaces:

  • Total dispatches (all-time, last 30d, last 7d)
  • Per-agent calibration table (Total / Clean / Blocked / Errored / Partial / Ratio)
  • Per-agent recent outcome trend (last 10 dispatches encoded as c7 b1 e0 p0)
  • ⚠ Repeat-blocker callouts (any agent with >=2 blocks in window)
  • Top 3 failing (agent, task) pairs

/agent-fitness --days 7 — same as above but windowed to last N days. Useful for "is this agent failing more lately?"

Computation

Calibration ratio per agent: clean / (total − errored). The denominator excludes hard errors (infrastructure/tool failures) and only counts agent-level outcomes.

Trend signal: any agent with >=2 blocks in the window surfaces a repeat-blocker callout. The classifier was hardened in v0.5.1 to catch real Lyra-style block patterns ("BLOCKED:", "are denied", "blocker-found" etc).

When to run

  • After major sessions where multiple sub-agents fired (today S30d had ~7 dispatches)
  • Periodically (weekly?) to spot trends
  • Before relying on a sub-agent for something heavy ("am I about to dispatch an agent that's been failing?")

What it does NOT do

  • Does not auto-reroute dispatches based on calibration. Surfaces data; principal decides.
  • Does not LLM-classify failures — heuristic only (keyword matching on result content).
  • Does not pool across sessions for cluster analysis. Future v0.5.x.
  • Does not include task-type clustering. All tasks pooled per agent.

Cross-links

  • [[concepts/Somatic v0.5.0 Agent Fitness]] — spec
  • [[memory/AGENT_FITNESS]] — the data
  • [[concepts/Somatic Roadmap]]
  • [[memory/TRUST_DECAY]] — sibling claim → outcome ledger

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free