Tune Agent

/tune-agent — Reflexion patch for an agent's prompt

/tune-agent <name> spawns the agent-tuner agent with collected evidence and proposes a focused patch to that agent's .md file.

When to invoke

• /mine-transcripts produced agent-tuning.md listing agents with ≥2 failures
• /eval shows a regression for a specific agent vs prior runs
• Verifier or critic flagged the same failure pattern from the same agent twice in this session
• The user says "this agent keeps doing X — fix it"

What you do

1. Resolve the agent. ~/.claude/agents/<name>.md must exist. If not, stop.

2. Collect evidence. Pull from:

   • Most recent ~/.claude/telemetry/mining-runs/*/agent-tuning.md (if applicable)
   • Verifier/critic outputs in this session that mention the agent
   • Any context the user just provided

   Need ≥2 pieces. If you have only 1, tell the user — wait for more before tuning.

3. Spawn agent-tuner. Use the Agent tool with subagent_type: agent-tuner. Pass the evidence as a structured prompt:

   Agent to tune: <name>
   Evidence:
   1. <task summary> — <what went wrong> — session=<id>
   2. <task summary> — <what went wrong> — session=<id>
   3. <task summary> — <what went wrong> — session=<id>
   
   Desired behavior: <if user specified, otherwise infer>
   

4. Read the tuner's output. It ends with one of:

   • RESULT_agent-tuner=PATCHED — surface the diff and rationale to the user; ask "apply or revert?"
   • RESULT_agent-tuner=INSUFFICIENT_EVIDENCE — tell user, suggest waiting
   • RESULT_agent-tuner=AMBIGUOUS — surface options to user
   • RESULT_agent-tuner=NO_PATCH_NEEDED — evidence didn't add up; explain

5. Apply or revert.

   • On user approval: leave the change in place (the tuner already wrote it).
   • On rejection: cp ~/.claude/agents/.history/<name>-<ISO>.md ~/.claude/agents/<name>.md to restore.
   • Either way, log to ~/.claude/agents/.history/CHANGELOG.md (the tuner already did this).

6. Re-eval. After applying, run /eval <name> to confirm the patch doesn't regress other behaviors.

Cost note

The tuner runs on Opus (precise prompt engineering matters). One invocation is typically <$0.50. If you find yourself tuning the same agent more than weekly, the agent's design is wrong — propose a redesign, not another patch.

Anti-patterns to refuse

• Tuning on a single occurrence (refuse — collect more evidence first)
• Tuning based on user vibes rather than logged failures
• Tuning multiple agents in one shot — each /tune-agent is one agent
• Skipping the post-patch /eval — you must verify the patch doesn't regress