self-eval

/self-eval — Self-Evaluation Command

You are the self-evaluation orchestrator. You evaluate the quality, necessity, and correctness of recent work using the ProductionOS Self-Eval Protocol.

Step 0: Preamble

Before executing, run the shared ProductionOS preamble (templates/PREAMBLE.md).

Step 1: Determine Evaluation Target

Based on $ARGUMENTS.target:

target = "last"

# Find the most recent agent output
ls -t .productionos/*.md .productionos/**/*.md 2>/dev/null | head -5

Read the most recent artifact. This is your evaluation target.

target = "session"

# Find all work done this session
ls -t .productionos/self-eval/ 2>/dev/null | head -20
git log --oneline --since="4 hours ago" 2>/dev/null | head -20
git diff --stat HEAD~10 2>/dev/null | tail -5

Evaluate ALL work produced this session. Produce a session-level summary.

target = "diff"

# Evaluate the current git diff
git diff --stat 2>/dev/null
git diff --name-only 2>/dev/null

Evaluate all uncommitted changes against the self-eval protocol.

target = specific path

Read the specified file and evaluate it.

Step 2: Dispatch Self-Evaluator Agent

Read agents/self-evaluator.md and dispatch:

Agent tool:
  description: "self-evaluator: Evaluate {target description}"
  prompt: "{self-evaluator role + instructions}\n\nTASK: Evaluate {target}\nDEPTH: $ARGUMENTS.depth\nOUTPUT: .productionos/self-eval/{timestamp}-eval.md"

Step 3: Process Results

Read the evaluation output. Based on score:

Score >= 8.0 — PASS

✅ Self-Eval PASS (X.X/10)
{summary of findings}
Logged to .productionos/self-eval/{file}

Score 6.0-7.9 — CONDITIONAL (self-heal if enabled)

If $ARGUMENTS.heal is "on":

1. Read the lowest-scoring questions
2. Generate targeted fix instructions
3. Dispatch the original agent (or self-healer) to address issues
4. Re-run self-eval (max 3 loops)
5. Report final result

If $ARGUMENTS.heal is "off":

⚠️ Self-Eval CONDITIONAL (X.X/10)
{issues that need attention}
Run /self-eval --heal on to attempt self-fix