hassan-mohiddin/freeflow / .deprecated/skill-tooling-pre-rewrite/skills/evaluate-skill
evaluate-skill
VerifiedUse when evaluating, comparing, or making readiness claims about agent-skill behavior.
references/grading-and-revision.md
View on GitHub# Grading And Revision
## Semantic Grading
Use a fresh context with criteria fixed before the subject run. Hide meaningful variant labels, randomize opaque labels when comparing runs, sanitize paths, and provide only the evidence needed for the assertion.
Return pass, fail, or uncertain with cited evidence. Keep uncertainty visible.
## Diagnose Before Editing
Classify the failed boundary:
- **Activation timing:** loaded too early, too late, or not at all.
- **Pre-load language:** description relied on vocabulary unavailable before the body.
- **First-read body:** job, terms, normal method, or stop conditions were not understandable from guaranteed context.
- **Dependency:** required base context was undeclared, undelivered, or supplied only by ambient installation.
- **Progressive disclosure:** normal-path guidance was hidden in a reference, or a resource lacked a recognizable read condition.
- **Nearby behavior:** activation hijacked a task that should have passed through safely.
- **Retained use:** later-turn behavior failed or the chosen evidence could not prove retention.
- **Instruction pressure:** wording, placement, or stop condition failed.
- **Case system:** fixture, prompt, host capability, isolation, or grader was wrong.
Do not add broad prose before identifying which boundary failed.
## Revision
Change one measured pressure point when possible:
- description cues;
- first-read wording;
- rule placement;
- dependency declaration;
- reference boundary;
- stop or exit condition;
- case or grader when evidence shows the evaluation is wrong.
Keep fixed criteria stable. If the criterion changes, record that the old run no longer answers the same question.
## Reruns
One invocation owns one complete case. Do not repeat one side, reuse a settled control, adopt partial evidence, or rerun merely to erase a saved failure.
After a measured revision or resolved infrastructure diagnosis, invoke the whole fixed case again. If variance evidence is required, represent repetitions as independent whole-case invocations and keep every result visible.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |