Build, run, and evolve Kevin's agent self-evaluation library. Use when the user says "benchmark yourself", "self improve", "agent eval library", "evaluate the agent", "make the judge harder", "Red Queen", "Gödel Machine", "eval the loop", "optimize this guardrail", or asks to turn an agent failure into a benchmark, receipt, rubric, or regression case.
Use this skill when an agent needs to benchmark its own behavior, not merely finish the task.
The local eval library lives at:
evals/agent-self-improvement/suite.json — benchmark cases and evaluator-evolution rulesscripts/run-agent-self-evals.ts — deterministic receipt scorerevals/agent-self-improvement/examples/ — example receipts and fixtureswiki/concepts/agent-self-improvement-eval-library.md — architecture and operating modelAn agent run should leave a small JSON receipt of what mattered:
{
"runId": "short-run-id",
"events": [{ "type": "qmd-search", "detail": "searched wiki first" }],
"commands": [{ "cmd": "npm run build-index", "exitCode": 0 }],
"filesTouched": ["wiki/tools/loopy.md"],
"promotedPages": ["wiki/tools/loopy.md"],
"notes": ["why this run changed the evaluator"]
}Then run:
npm run agent-self-eval
npm run agent-self-eval:test
npm run agent-self-eval -- --receipt evals/agent-self-improvement/examples/passing-receipt.jsonThe first command validates the suite. The second scores a run receipt.
npm run agent-self-eval -- --receipt <path>.Skill changes and evaluator changes are different control loops. For a SKILL.md candidate, keep the evaluator fixed and apply the skillOptimization contract in evals/agent-self-improvement/suite.json:
Route authoring and optimization through skill-creator. Use Red Queen evaluator evolution only when the fixed judge itself has become the bottleneck; never change the judge and the skill in the same comparison.
When the candidate is a guardrail policy, use a paired objective and change only that policy artifact:
Use SantanderAI/autoguardrails@1ca0c9b as a compact reference implementation, not as a production safety benchmark. Its bundled perfect score uses a deterministic stub and a single-turn fixed suite.
Before adding a case, decide what is actually being graded. Use wiki/tools/awesome-evals.md as the source map and prefer the narrowest grader that proves the behavior.
| Object being graded | Default grader | Use when |
|---|---|---|
| Structured final output | deterministic assertions | JSON shape, citations, links, exact labels, policy fields, build artifacts |
| Subjective final output | binary LLM judge after error analysis | taste, usefulness, tone, critique quality, when deterministic checks are insufficient |
| Tool trajectory | trajectory match plus trace review | known workflows where the expected tool sequence matters |
| Agent outcome | environment or state diff | DB/files/API state changed; the transcript can lie |
| Harness quality | run receipt plus trace spans | tool choice, context loading, approvals, costs, closeout proof |
| Benchmark integrity | contamination, label-error, and saturation checks | comparing models or publishing capability claims |
| Self-improving agent | evolving evaluator with held-out negatives | the agent can learn to satisfy a fixed judge |
Concrete rules from the current eval source map:
pass@k capability from pass^k reliability before reporting numbers.A stable judge becomes a target, but a judge that changes during a run destroys the meaning of improvement. Use controlled utility evolution, not continuous rubric mutation:
The judge may improve alongside the agent; it does not automatically deserve promotion. A weak or biased anchor remains a weak floor, and the Red Queen Gödel Machine paper provides only preliminary, epoch-local guarantees—not global convergence. The local machine-readable contract lives under evaluatorEvolution in evals/agent-self-improvement/suite.json; its negative invariants run through npm run agent-self-eval:test. [Source: arXiv:2606.26294v2, methods, experimental design, limitations; CaMLSys explainer, 2026-07-19; replayed 2026-08-11]
loopy — craft bounded agent loops and loop skills.x-bookmark-absorb — promote bookmark evidence into pages, skills, tools, playbooks, workflows, diagrams, and durable objects.skill-creator — create or prune executable skills after an eval exposes a repeated procedure.agent-iteration-loop — implementation loop that should emit receipts for substantial work.