Skip to content

agentsope/SkillAlchemy / skills/agentsop-metric-design

agentsop-metric-design

Verified

Decomposed, multi-criteria metric design for LLM pipelines. The metric IS the model — change the metric and the optimizer changes behavior. Decompose by default; bool during compile, float during eval; calibrate against human; mitigate judge bias. Search keywords: LLM-as-judge, llm as judge, eval metric, evaluation score, scoring function, rubric, RAGAS, G-Eval, judge bias, verbosity bias, how to evaluate LLM output.

FileKindSize
README.mdmarkdown2.4 KB
SKILL.mdskill md25.3 KB
intermediate/operation_candidates.jsonconfig6.6 KB
references/R1-source-evidence.mdmarkdown10.1 KB
references/R2-judge-bias-catalog.mdmarkdown7.9 KB
# metric-design skill

A Phase-D **tool skill** for designing the metric an LLM optimizer maximizes. The
core idea: **the metric IS the model** — a DSPy/GEPA/MIPRO optimizer is a
black-box search that chases whatever the metric rewards, so a bad metric beats a
good optimizer every time. This skill decomposes a holistic LLM-as-judge into 3-6
orthogonal yes/no sub-judges, returns `bool` during compile and `float` during
eval (`trace is not None`), bakes length in as a deterministic scalar (judges
over-prefer length), uses a cross-family judge to dodge self-preference,
runs a bias-probe suite (length / self-preference / position / rubric-order), and
**human-calibrates on >=20 spot-checks before any compile** — producing a metric
function plus a calibration receipt that DSPy compilers, the LlamaIndex
Faithfulness/Relevancy/Retriever triad, RAGAS, and TruLens consume.

**Version**: 0.1.0 - **Phase**: D - **Tier**: core - **Status**: opinionated

## Files

- `SKILL.md` — main skill (7 sections: 何时激活 / 核心心智模型 / SOP / 操作模型 /
  困境决策案例 / 反模式与边界 / 跨框架对照)
- `references/R1-source-evidence.md` — each load-bearing claim traced to its
  source (dspy-sop Case C + ops op-004/op-024/op-030; llamaindex-sop OP-10
  triad; judge-bias papers) with verbatim quotes.
- `references/R2-judge-bias-catalog.md` — catalog of LLM-as-judge biases
  (length, self-preference, positional, rubric-order, score-ID, hedging) with a
  detection probe and mitigation op per bias.
- `intermediate/operation_candidates.json` — the 10 operations (OP-M01…OP-M10)
  in id / name / trigger / action / output / evidence form.

## Cross-links

- `[[agentsop-dspy]]` — the optimizer side; consumes this metric (Case C, op-030).
- `[[agentsop-llamaindex]]` — RAG evaluator triad reused as sub-judges (OP-M06).
- `[[agentsop-domain-eval-set]]` — defines *what* is scored; this skill defines *how*.
- `[[agentsop-regression-gate]]` — consumes the calibrated metric as the per-PR gate.

## Source basis

- DSPy SOP — metric mode switch, decomposition, calibration, GEPA pairing
  (`output/dspy-sop-skill/`) [dspy.ai/learn/evaluation/metrics/]
- LlamaIndex SOP — Faithfulness/Relevancy/Retriever evaluator triad, OP-10
  (`output/llamaindex-sop-skill/SKILL.md`)
  [developers.llamaindex.ai/python/framework-api-reference/evaluation/]
- Judge-bias research — [arxiv.org/pdf/2506.02592], [arxiv.org/pdf/2509.26072];
  GEPA [arxiv.org/abs/2507.19457]
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall