Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/joymin5655/Agentnpx agentmods add skills/joymin5655/agent/harness-auditWrote 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/joymin5655/agent/harness-audit)<a href="https://agentmods.dev/skills/joymin5655/agent/harness-audit"><img src="https://agentmods.dev/badge/skills/joymin5655/agent/harness-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/skills/joymin5655/agent/harness-audit"><img src="https://agentmods.dev/badge/skills/joymin5655/agent/harness-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 | $0.00105 | $0.02352 |
| Opus 5.5 | $0.00042 | $0.00941 |
| Sonnet 5.5 | $0.00021 | $0.00470 |
| Haiku 4.5 | $0.00011 | $0.00235 |
Grade A, and why
harness-audit 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 9d 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 — 176 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/harness-audit
Goal
Produce one interpreted health report of the harness in a single read-only
dry-run. The machine integrity layer (P1-1 doc-reality.sh, the CI jobs, and the
verify-all.sh runner) is the enforcement; this skill is the consumer that
runs it once, reads off a per-check pass/fail table, and turns each failure into a
diagnosis a maintainer can act on.
Run it from the root of a harness checkout: it audits that checkout's own gates.
Concretely the report answers three questions with evidence:
- Is the harness passing its own gates right now? — a per-check PASS / FAIL /
SKIP table, taken verbatim from
verify-all.sh. - What does the doc-reality gate (P1-1) say? — cited explicitly, because it is the gate that catches the harness lying about itself (phantom paths + drifted counts), and it is the one item most easily missed in a wall of green.
- For every non-PASS: why, and what next? — the check's purpose, the likely root cause, the recommended fix, and whether it warrants a new backlog row.
This skill supplies the interpretation; it does not re-implement any check.
The checks live in core/tests/ and CI; here you run them and read the results.
Why this is a layer, not a duplicate of P1-1
P1-1 (core/tests/doc-reality.sh) is a machine gate: it runs in CI, exits 0/1,
and mechanically blocks doc drift. This skill is the agent consumer that sits on
top of that gate — it invokes the whole machine layer in one dry-run and interprets
the results (root-cause, fix, backlog follow-up) the way a maintainer would. It reads
doc-reality's verdict; it never re-derives phantom-path or count logic. Deleting this
skill would not weaken any gate (CI still enforces them); deleting doc-reality would.
That asymmetry is the layer relationship: the gate enforces, the audit interprets.
This mirrors the machine-gate-vs-agent-consumer split the plan draws for H-2.
Steps
1. Run the machine integrity layer (one dry-run)
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.
- 9d ago Changed · +4 lines 2c4d697345d8
- 14d ago Changed · +2 lines eae0b29b6065
- 1mo ago Changed 67e898e0280c
- 2mo ago First seen · 170 lines · 105 tokens per session scan A 701683c8b989
harness-audit is a skill published in the GitHub repository joymin5655/Agent (2 stars, last pushed yesterday), licensed MIT. It adds 105 tokens to every session and 2,352 once invoked, about $0.0004 per session on Opus 5.5. 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.
harness-audit compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| harness-auditjoymin5655/Agent | Skill | 2 | yesterday | MIT | A | 105 | Claude Code |
| agent-prompt-auditavibe-bot/avibe | Skill | 637 | yesterday | MIT | A | 105 | Codex |
| computer-useavibe-bot/avibe | Skill | 637 | yesterday | MIT | A | 19 | Claude Code, Codex |
| background-watch-hookavibe-bot/avibe | Skill | 637 | yesterday | MIT | A | 32 | Claude Code, Codex |
| use-avibe-harnessavibe-bot/avibe | Skill | 637 | yesterday | MIT | A | 36 | Claude Code, Codex |
| use-avibe-vaultavibe-bot/avibe | Skill | 637 | yesterday | MIT | A | 36 | Claude Code, Codex |
| use-show-pagesavibe-bot/avibe | Skill | 637 | yesterday | MIT | A | 39 | Claude Code, Codex |