76 tagged eval, measured the same way as everything else here.
Browse within: skills 58harness-engineering 55sdd 55spec 55LLM 18AI Safety 8prompt-engineering 8skill 6Evaluation 5agent 5
Skill Codex
A Comet workflow step for turning an approved design into an implementation plan and completed code changes. Comet is the project workflow system described by the add-on.
Skill Claude CodeCodex needs its repo
Authors deterministic, LLM rubric, and human graders for skillgrade evaluations. Use when creating scoring scripts, writing evaluation rubrics, setting up human review workflows, or combining multiple graders with weighted scoring. Don't use for setting up eval pipelines, configuring eval.yaml defaults, or general…
Skill Claude Code
Review a change in the EvalBench repo for (a) does it actually work — verified by running the tests and style checks, (b) does it follow EvalBench architecture — base-class contracts, config-key registration, PYTHONPATH-relative imports, sandbox isolation, concurrency safety, docs, (c) does it still build and deploy …
Skill Claude CodeCodex ✓ vendor
Stage 3 of Clinical ASR Flywheel. Score a NeMo manifest, produce the five-section KER leaderboard (by-ipasource diagnostic). Not for ASR auth (/riva-asr).
Skill Claude CodeCodex
Improve AI application with evaluation-driven development. Define eval criteria, instrument the application, build golden datasets, observe and evaluate application runs, analyze results, and produce a concrete action plan for improvements. ALWAYS USE THIS SKILL when the user asks to set up QA, add tests, add evals…
Skill Claude Code
Use this skill when the user wants to commit, write a commit message, or save their staged or current changes with a clear message to git history.
Skill Claude Code
Build, lint, and evaluate agent skills that demonstrably change behaviour, instead of writing adjectives that change nothing. Use whenever the user asks to create a skill, write a SKILL.md, improve or fix an existing skill, make a skill "actually work", or evaluate whether a skill does anything. Also use when the user…
Skill Claude CodeCodex
Before finalizing structured or extracted output in a domain with declared safety invariants, verify the output against your invariant suite and refuse to ship when a locked field was model-auto-filled. Invoke when the user or task involves high-stakes structured extraction, compliance fields, or pass/fail results…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: