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 agentmods add commands/studiokjm/ai-harness-template/evaluategit clone --depth 1 https://github.com/studioKjm/ai-harness-templateWrote 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/commands/studiokjm/ai-harness-template/evaluate)<a href="https://agentmods.dev/commands/studiokjm/ai-harness-template/evaluate"><img src="https://agentmods.dev/badge/commands/studiokjm/ai-harness-template/evaluate.svg" alt="Measured on agentmods" 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 | $0.00040 | $0.01628 |
| Opus 5 | $0.00020 | $0.00814 |
| Sonnet 5 | $0.00008 | $0.00326 |
| Haiku 4.5 | $0.00004 | $0.00163 |
Grade A, and why
evaluate 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.
How it starts
The opening of the file, as written. The whole thing — 204 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/evaluate — 3-Stage Verification
구현 결과를 3단계로 검증한다
Instructions
You are now the Evaluator agent. Verify the implementation against the seed spec.
Phase 0: State Audit (FIRST STEP)
- Locate seed spec —
.harness/ouroboros/seeds/seed-v*.yaml(latest)- If none → abort: "No seed to evaluate against. Run /interview → /seed first."
- Check for prior evaluations —
.harness/ouroboros/evaluations/- If recent (<1h) PASS with no code changes → skip re-evaluation
- If recent FAIL → surface prior findings; focus on whether they were addressed
- Detect scope — which files changed since last commit? (
git diff --stat)- Narrow evaluation to changed files when possible
Subagent Delegation
검증의 정확도와 속도를 높이기 위해 subagent를 활용합니다:
Main Agent (Evaluator)
├─ Subagent → Stage 1 (Mechanical): 게이트 실행을 별도 에이전트에 위임
│ └ .harness/detect-violations.sh 실행 + 결과 보고
├─ Main → Stage 2 (Semantic): 시드 대비 AC/목표/제약 직접 검증
└─ Main → Stage 3 (Judgment): 코드 품질 판단
Claude Code에서 subagent 사용:
- Stage 1의 기계적 검증은
Agent도구로 별도 subagent에 위임 가능 - subagent가 게이트 스크립트를 실행하고 결과만 반환
- 메인 에이전트는 Stage 2/3에 집중하여 병렬 처리 효과
Stage 1: Mechanical Verification ($0 cost)
Run automated checks — these cost nothing and catch obvious issues:
# 1. Harness gates
.harness/detect-violations.sh
# 2. Layer separation check
.harness/gates/check-layers.sh
# 3. Lint (if available)
# TypeScript: npx eslint . --quiet
# Python: ruff check . || python -m flake8
# 4. Type check (if available)
# TypeScript: npx tsc --noEmit
# Python: mypy . || pyright
# 5. Build (if available)
# Next.js: npm run build
# Python: python -m py_compile
# 6. Tests (if available)
# npm test || pytest
Report format:
═══ Stage 1: Mechanical ═══════════════════════
Harness Gates: PASS | FAIL
Layer Check: PASS | FAIL | SKIP
Lint: PASS | FAIL | SKIP (not configured)
Type Check: PASS | FAIL | SKIP
Build: PASS | FAIL | SKIP
Tests: PASS | FAIL | SKIP
─────────────────────────────
Result: PASS | FAIL
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.
- 4d ago First seen · 204 lines · 40 tokens per session scan A dd21aef22b6c
evaluate is a command published in the GitHub repository studioKjm/ai-harness-template (43 stars, last pushed 3mo ago), licensed MIT. It adds 40 tokens to every session and 1,628 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 2026-08-30.
Other commands, from other repositories
VibeGuard: ExecPlan
Long-term task execution plan — generates self-contained execution documents from SPEC, supports cross-session recovery.
VibeGuard: Review
Structured code review - first run the guard to obtain the baseline, then review according to security → logic → quality → performance priority.
VibeGuard: Live Truth
Verify live claims with fresh facts, inferences, and unresolved gaps.
capture-feedback
Quick feedback capture with structured signals.
check-gates
Run a Pre-Action Gate check against prevention rules before executing a risky action.
harness-adopt
Apply Harness Starter Kit to the current target repository with the prompt-first adoption workflow.