Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add leee880619-commits/ClaudeCode-Harness-Setup-Assistant/plugin install harness-architectWrote 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/leee880619-commits/claudecode-harness-setup-assistant/audit)<a href="https://agentmods.dev/commands/leee880619-commits/claudecode-harness-setup-assistant/audit"><img src="https://agentmods.dev/badge/commands/leee880619-commits/claudecode-harness-setup-assistant/audit/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/commands/leee880619-commits/claudecode-harness-setup-assistant/audit"><img src="https://agentmods.dev/badge/commands/leee880619-commits/claudecode-harness-setup-assistant/audit.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 | $0.00051 | $0.06811 |
| Opus 5 | $0.00026 | $0.03406 |
| Sonnet 5 | $0.00010 | $0.01362 |
| Haiku 4.5 | $0.00005 | $0.00681 |
Grade A, and why
audit scanned grade A with 1 finding 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 8d 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.
Reads agent configuration directorieslowAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
- `{대상}/.claude/settings.json` Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 345 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Harness Unified Audit
기존 Claude Code 하네스에 대해 3개 감사를 병렬 실행하고 단일 통합 보고서를 제공합니다. 시나리오 B(기존 환경 점검) 의 표준 엔트리입니다.
통합 감사 구성
| 감사 | 근본 질문 | 출력 등급 |
|---|---|---|
harness-auditor |
"파일 구조가 올바른가" | CRITICAL / HIGH / MEDIUM / LOW |
ops-auditor |
"실행할 때 실패하는가" | RISK-HIGH / RISK-MED / RISK-LOW |
fit-auditor |
"이 프로젝트에 여전히 맞는가" | MAJOR-DRIFT-CRITICAL / MAJOR-DRIFT-MED / MINOR-DRIFT / ALIGN |
세 감사는 모두 read-only 이며 파일을 수정하지 않습니다. 감사 결과를 대상 프로젝트에 쓰지 않습니다.
범위
- 이 커맨드는 감사 3종(harness / ops / fit) 을 병렬 실행하는 유일한 슬래시 엔트리 입니다. 개별 auditor 를 단독 슬래시 커맨드로 호출하는 경로는 autocomplete 에 노출되지 않습니다 (혼란 방지). 단독 실행이 필요할 때는 아래 "개별 auditor 단독 실행" 섹션을 참조하세요.
- 대규모 하네스 (에이전트 30+ 개, CLAUDE.md 여러 파일 @import) 에서 통합 보고서가 커지면 "보고서 크기 관리" 섹션의 자동 축약 규칙이 적용됩니다.
Pre-flight Gate
커맨드 진입 시 사용자에게 먼저 안내 (AskUserQuestion 없이 텍스트):
이 커맨드는 진단 전용입니다. 3개 감사가 발견한 문제를 자동으로 수정하지 않습니다. 보고서 출력 후, 마지막 섹션의 "🛠 다음 실행 방법 (How to apply)" 가이드에 따라
/harness-architect:harness-setup <대상 경로>를 재실행하여 개선을 적용하거나 수동 편집을 수행합니다.
이후:
-
$ARGUMENTS또는 AskUserQuestion 으로 대상 프로젝트 절대 경로 수집. 경로 유효성 검증 (test -d). -
하네스 최소 존재 검증 — 다음 중 하나라도 존재해야 감사 진행:
{대상}/CLAUDE.md{대상}/.claude/settings.json{대상}/.claude/agents/(파일 1개 이상){대상}/playbooks/(파일 1개 이상)
-
모두 미존재 시 AskUserQuestion 분기:
먼저 harness-setup 실행—/harness-architect:harness-setup안내 후 종료취소— 종료
-
User Work Signal Gate — 감사 대상(사용자 작업 파일)이 1개 이상 존재하는지 단일 이진 신호로 확인. 이 gate 는 임계치·경과일·아키타입 분기 없이 동작하며, 통과 시 기존 감사 플로우와 완전히 동일하게 진행됩니다.
TARGET="{대상 절대 경로}" # 공백·한글 경로 대응: 반드시 변수화 + 따옴표 인용 USER_WORK=$(find "$TARGET" -type f -size +0c \ -not -path "$TARGET/.claude/*" \ -not -path "$TARGET/.cursor/*" \ -not -path "$TARGET/.vscode/*" \ -not -path "$TARGET/.idea/*" \ -not -path "$TARGET/docs/*" \ -not -path "$TARGET/.git/*" \ -not -path "$TARGET/node_modules/*" \ -not -path "$TARGET/.venv/*" \ -not -path "$TARGET/venv/*" \ -not -path "$TARGET/dist/*" \ -not -path "$TARGET/build/*" \ -not -path "$TARGET/.next/*" \ -not -path "$TARGET/target/*" \ -not -path "$TARGET/__pycache__/*" \ -not -path "$TARGET/.pytest_cache/*" \ -not -path "$TARGET/.mypy_cache/*" \ -not -name 'README*' -not -name 'LICENSE*' \ -not -name '.gitignore' -not -name '.editorconfig' \ -not -name '.gitattributes' -not -name '.gitmodules' \ -not -name '.DS_Store' -not -name 'Thumbs.db' \ -not -name 'package-lock.json' -not -name 'yarn.lock' \ -not -name 'pnpm-lock.yaml' -not -name 'uv.lock' \ -not -name 'Cargo.lock' -not -name 'go.sum' \ -not -name 'poetry.lock' -not -name 'Pipfile.lock' \ 2>/dev/null | head -1)
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.
- 8d ago First seen · 345 lines · 51 tokens per session scan A 9d53de6278a5
audit is a command published in the GitHub repository leee880619-commits/ClaudeCode-Harness-Setup-Assistant (2 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 51 tokens to every session and 6,811 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
ai-incident
Registrar y analizar incidentes donde recomendaciones de Savia fueron incorrectas.
arch-health
Salud arquitectónica — fitness functions, drift detection, coupling metrics.
get-flow-run
Inspect a Postman Flow run by its Run ID — per-block logs, failing block, and status.
t800-fix
A command for making a narrow, predefined fix to existing agents, skills, commands, rules, or hooks. A fix-pack is a file listing the exact changes to make.
debug
Systematic root-cause investigation, then fix with TDD — never proposes a fix before finding the cause; resumable diagnosis + regression test in .orc/ state. Workspace-aware. For a quick one-off look without session state, prefer the bundled /debug.
t800-doctor
A command that runs a scripts-only health report for a T-800 or target plugin setup. It reports the version, profile, memory location, state, item counts, always-applied items, and recent audits.