memradar: Skill for Claude Code

.claude/skills/generate-eval-samples/SKILL.md

generate-eval-samples is a skill for Claude Code from on1659/memradar. It costs 0 tokens per session (1,226 once invoked), scanned A, original, MIT.

A script for creating synthetic test examples to measure how accurately an AI role classifier assigns categories to usage patterns.

In plain words
What is it for?
Use it to generate up to 109 evaluation samples with Z.AI and optionally run the classifier evaluation afterward.
Why use it?
It makes it possible to evaluate the classifier with a repeatable set of varied examples and identify weak results.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: names the AskUserQuestion tool; positional $N argument.

This is on1659/memradar's own configuration. It tells Claude Code how to work on memradar itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything memradar configures →

Reuse

Borrowing it

Nothing to install: this file belongs to on1659/memradar. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/on1659/memradar/master/.claude/skills/generate-eval-samples/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/on1659/memradar

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for generate-eval-samples

README.md
[![agentmods](https://agentmods.dev/badge/skills/on1659/memradar/generate-eval-samples.svg)](https://agentmods.dev/skills/on1659/memradar/generate-eval-samples)
Your own site
<a href="https://agentmods.dev/skills/on1659/memradar/generate-eval-samples"><img src="https://agentmods.dev/badge/skills/on1659/memradar/generate-eval-samples.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,226 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00000 $0.01226
Opus 5 $0.00000 $0.00613
Sonnet 5 $0.00000 $0.00245
Haiku 4.5 $0.00000 $0.00123

Measured 4d ago against content hash ac5538ea5d18, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

generate-eval-samples scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/generate-eval-samples/SKILL.md · 105 lines

How it starts

The opening of the file, as written. The whole thing — 105 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Generate AI Role Evaluation Samples

AI 역할 분류기(src/lib/usageProfile.ts)의 정확도를 측정하기 위한 합성 평가 샘플 109개를 Z.AI (Zhipu GLM-4.6) 로 생성한다.

사용법

인자 (optional)

  • --model <name>: Z.AI 모델 지정 (기본: glm-4.6). 예: glm-4.5-air, glm-4-plus
  • --skip-eval: 생성만 하고 평가는 스킵
  • --clean: 기존 tests/fixtures/role-eval-samples/ 비우고 새로 생성
  • 그 외 토큰은 무시

실행 전 점검

  1. Z.AI 키 확인 — 다음 중 하나 필수:

    • .env 파일에 ZAI_API_KEY=... (권장, .gitignore 처리됨)
    • 또는 환경변수 export ZAI_API_KEY="..."
    • 없으면 STOP 하고 사용자에게 .env.example 복사 안내:
      cp .env.example .env
      # .env 에 키 입력
      
    • 키 발급: https://z.ai/
  2. 모델 확인 — 기본 glm-4.6 (최신 플래그십). 변경하려면:

    • .envZAI_MODEL=glm-4.5-air 같이 지정
    • 사용자가 GLM 5.x 등 새 모델 언급하면 그 이름 그대로 사용 (스크립트가 거르지 않음)
  3. 작업 트리 체크tests/fixtures/role-eval-samples/ 기존 파일 확인

    • 파일 있으면 덮어쓰기 여부 AskUserQuestion 으로 확인
    • --clean 있으면 자동 삭제
  4. 네트워크 확인 — 중국 외부 접속 때 Z.AI 가끔 느림. 15분 타임아웃.

실행 플로우

  1. 생성 (약 10-20분, 109 × API 호출):

    npx tsx scripts/generate-eval-samples-zai.mts
    
    • CONCURRENCY=4 로 병렬 호출
    • 실패 샘플은 재시도 3회 후 폐기
    • 결과: tests/fixtures/role-eval-samples/sample-*.json (최대 109개)
  2. 평가 (--skip-eval 아니면 자동 실행):

    npx tsx scripts/eval-role-samples.mts
    
    • 각 샘플 → analyzeUsageTopCategories 실행
    • top1 역할이 acceptableRoles 포함되는지 확인
    • 결과: docs/AI-ROLE-EVAL-RESULTS.md
  3. 요약 출력:

    • 전체 정확도
    • 카테고리별 / 난이도별 정확도
    • Top 5 오분류 쌍
    • 생성 성공률 (예: 107/109 = 98.2%)

비용 추정

  • GLM-4.6 (flagship): 입력 $0.60/1M, 출력 $2.20/1M
  • 109 세션 × 평균 200 msgs ≒ 22k input + 35k output tokens
  • 총 약 $0.10-0.15 (Haiku 보다 저렴)

결과 해석 가이드

>80% 정확도

Phase 1/2 로직 건강. 튜닝 불필요.

50-80% 정확도

혼동행렬 확인. 특정 카테고리 keyword dict 약함 의심.

<50% 정확도

  • GLM-4.6 한국어 bias 가능성 (generator 필드로 추적)
  • 또는 Phase 1/2 로직 근본 문제

결과 발표 시 generator: glm-4.6 명시. 다른 LLM 으로 재생성해서 cross-check 하는 게 ideal.

안전 규칙

  • 절대 실제 유저 세션 파일 건드리지 않음 — 오직 tests/fixtures/role-eval-samples/ 만 읽고 씀
  • 샘플 내용에 실제 회사/사람 이름 있으면 플래그 (프롬프트에서 금지했지만 누락 가능)
  • .gitignore 처리된 폴더라 git add 금지

Read the full file on GitHub · 105 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 105 lines · 0 tokens per session scan A ac5538ea5d18

Subscribe to this mod's changes

generate-eval-samples is a skill published in the GitHub repository on1659/memradar (11 stars, last pushed 5d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,226 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-02.

Related

Other skills, from other repositories

cli-eval

Create and run evaluation suites, watch live benchmark progress, view scorecards, compare model performance, and integrate eval runs with CI workflows from the CLI.

diegosouzapw/OmniRoute · 34 tokens

model-merging

Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task…

davila7/claude-code-templates · 73 tokens

darwinian-evolver

Evolve prompts/regex/SQL/code with Imbue's evolution loop.

NousResearch/hermes-agent · 22 tokens

validate

Validate Semantica pipelines, extraction quality, graph schemas, and ontology consistency. Returns structured error/warning checklists. Uses PipelineValidator, PipelineBuilder.validatepipeline(), GraphValidator, and OntologyValidator. Sub-commands: pipeline, step, dependencies, extraction, graph, ontology, performance.

semantica-agi/semantica · 0 tokens

launching-evals

Run, monitor, analyze, and debug LLM evaluations via nemo-evaluator-launcher. Covers running evaluations, checking status and live progress, debugging failed runs, exporting artifacts and logs, and analyzing results. ALWAYS triggers on mentions of running evaluations, checking progress, debugging failed evals…

NVIDIA/Model-Optimizer · 115 tokens

nemo-automodel-recipe-development

Create and modify NeMo AutoModel training and evaluation recipes, including YAML structure, builders, and execution flow.

NVIDIA/skills · 31 tokens