Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add bunhine0452/Ocul-PM --skill run-evalsgit clone --depth 1 https://github.com/bunhine0452/Ocul-PMWrote 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.
[](https://agentmods.dev/skills/bunhine0452/ocul-pm/run-evals)<a href="https://agentmods.dev/skills/bunhine0452/ocul-pm/run-evals"><img src="https://agentmods.dev/badge/skills/bunhine0452/ocul-pm/run-evals/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bunhine0452/ocul-pm/run-evals"><img src="https://agentmods.dev/badge/skills/bunhine0452/ocul-pm/run-evals.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00046 | $0.00700 |
| Opus 5 | $0.00023 | $0.00350 |
| Sonnet 5 | $0.00009 | $0.00140 |
| Haiku 4.5 | $0.00005 | $0.00070 |
Grade A, and why
run-evals 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 10d 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.
What it actually says
run-evals — EVALS.md 평가 실행
프로젝트 루트의 EVALS.md 가 이 프로젝트의 완료 정의(definition of done) 입니다.
절차
EVALS.md를 읽고 평가 항목을 파악한다. 파일이 없으면 사용자에게 만들지 물어본다. 항목은 두 종류로 구분해 다룬다:- capability — 새로 되게 만든 것 (이번 작업의 목표)
- regression — 전부터 되던 것 (깨지면 안 되는 것) — capability 만 돌리고 regression 을 건너뛰지 않는다.
- 각 항목을 실제로 실행/재현해 판정한다 — 코드를 읽고 "될 것 같다"로 판정하지 않는다. 판정 방식은 항목 성격에 맞게:
- 결정적 체크(명령 exit code·grep)가 1순위 — 가능하면 이것으로.
- 루브릭 채점(1~5)은 결정적 체크가 불가능한 품질 항목만 — 점수마다 근거 한 줄 필수. 4점 이상이 통과 (합격선 고정 — 세션마다 흔들리면 추이가 무의미).
- 사람 확인 필요 항목은 N/M 분모에서 제외하고 메모에 "사람 확인 대기 K건"으로 표기한다 (분모에 넣으면 확인 후에도 표에 실패로 영구히 남는다).
- 결과를
EVALS.md의## 기록표에 한 줄 append 한다 (표가 없으면 아래 형식으로 만든다):
| 날짜 | 스위트 | 통과 | 메모 |
|---|---|---|---|
| YYYY-MM-DD | 스위트명 | 통과수/전체수 | 실패 요약 |
- 실패 항목은 원인을 조사해 보고하고, 수정 후 재실행한다. 불안정한 항목은 연속 2~3회 통과를 기준으로 삼는다.
규칙
## 기록표의 형식(날짜 · 스위트 · N/M)은 바꾸지 않는다 — ocul-pm 회고 화면이 이 표를 읽어 추이를 그린다.- 통과율을 부풀리지 않는다. 애매하면 실패로 센다.
- 평가를 통과시키기 위해 평가 기준을 약화시키지 않는다 — 기준을 바꿔야 하면 이유를 보고하고 승인 받는다.
- 베이스라인 없는 수치를 만들지 않는다 — "N% 개선/절감"은 실측 베이스라인이 표에 있을 때만 말한다. 비교 대상이 기록된 적 없으면 절대값(N/M)만 보고한다.
- 가능하면 구현 전에 평가 항목부터 정의한다 (project-inception 이 만든 EVALS.md 가 그 출발점) — 구현 후에 끼워 맞춘 평가는 약하다.
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.
- 10d ago First seen · 34 lines · 46 tokens per session scan A b4a750b0a39d
run-evals is a skill published in the GitHub repository bunhine0452/Ocul-PM (7 stars, last pushed 2d ago), licensed MIT. It adds 46 tokens to every session and 700 once invoked, about $0.0002 per session on Opus 5. 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 2026-08-31.
Other skills, from other repositories
agentbro-pr-merge
Use when reviewing, fixing CI for, approving workflows for, or merging AgentBro pull requests into dev/main, especially external contributor PRs where contributor attribution matters.
test-router
A testing skill that chooses which checks to run based on the type, scope, and risk of a change.
dev-testing
A testing guide that defines when to use unit, integration, API, and end-to-end tests. Unit tests check small pieces of code, while end-to-end tests check a full user flow.
release-verification
A release-verification checklist for software versions and packages. It covers changelogs, tests, packaging, installation smoke checks, version-control actions, publishing, and checks after release.
host-contract-verification
A verification procedure for the rules between a host system and its hooks, command-line tools, startup scripts, workspace checks, visible responses, and fallback connections to MCP servers.
api-verification
An API verification workflow that creates both an .http request file and a .cjs JavaScript file, then runs an automation script to test them. An API is an interface through which software exchanges requests and responses.