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 Chemaclass/agnostic-ai --skill target-auditgit clone --depth 1 https://github.com/Chemaclass/agnostic-aiWrote 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/chemaclass/agnostic-ai/target-audit)<a href="https://agentmods.dev/skills/chemaclass/agnostic-ai/target-audit"><img src="https://agentmods.dev/badge/skills/chemaclass/agnostic-ai/target-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/chemaclass/agnostic-ai/target-audit"><img src="https://agentmods.dev/badge/skills/chemaclass/agnostic-ai/target-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00044 | $0.04245 |
| Opus 5 | $0.00022 | $0.02122 |
| Sonnet 5 | $0.00009 | $0.00849 |
| Haiku 4.5 | $0.00004 | $0.00424 |
Grade A, and why
target-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 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
outage. One curl would have saved all of it. How it starts
The opening of the file, as written. The whole thing — 357 lines — stays where its author put it; the contents beside it link to each section on GitHub.
target-audit
Answers one question: do these tools still read what agnostic-ai emits?
Adapters encode a snapshot of every vendor's config format. Those vendors
ship weekly. Drift is silent. A moved skills directory keeps syncing
cleanly, keeps sync --check green, and stops reaching the tool. This
skill finds that drift on a schedule instead of via a user bug report.
Arguments
- no args: audit every registered target and file an issue per confirmed finding. Filing is the default, not an opt-in.
claude zed kilo: audit only those--no-file-issues: stop at the report. Use it for a scratch run you do not intend to act on.--fix: file issues, then open a PR per fix bucket. Never merges.
Why filing is the default. It used to be opt-in, and two consecutive
runs proved that wrong. The 2026-08-19 run found 52 findings and filed
none. The 2026-08-27 run found 54 and filed none until a human asked, by
which point it was re-verifying the previous run's backlog instead of
auditing. The report lands in gitignored local/, so a run that does not
file leaves nothing behind: no tracked work, no dedupe set for the
next run, and a fresh re-derivation of the same findings a week later.
A report nobody can act on is not a cheaper audit, it is a wasted one.
Phase 1: Scope
scripts/target-facts.sh --list # registered targets
gh issue list --label target-audit --state all --limit 100 --json number,title,state,createdAt
Dedupe set. A finding already filed and open is not a new finding.
Note it as still-open #N and move on. A finding already filed and
closed that reappears is a regression. Say so loudly: it means a fix
was reverted or the vendor moved back.
Since when. Auditors need a date to bound their changelog reading.
Take the newest of: the most recent report under local/target-audit/,
the newest target-audit issue's createdAt, or the last commit touching
references/sources.md. All three under-report, because a clean run
leaves no trace anywhere, so the window is always a little wide. That is
the safe direction. Re-reading a changelog entry costs nothing, missing
one costs a release.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 357 lines · 44 tokens per session scan A d663134ffca7
target-audit is a skill published in the GitHub repository Chemaclass/agnostic-ai (12 stars, last pushed today), licensed MIT. It adds 44 tokens to every session and 4,245 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
eg-fix-bug
Fix a bug with Codex using problem doc, independent diagnosis, failing test, fix, review, and validation.
code_explorer
Explores the repository to locate primary source files, coupled UI components, and test files for bug reports or feature requests.
ios-simulator
Verify and debug native, React Native, Expo, or Flutter apps on an iOS Simulator with agent-device. Use when an agent needs to launch an app, inspect its live UI, tap, type, scroll, validate a code change, collect failure evidence, or reproduce a workflow on an iPhone or iPad Simulator.
cvm-ai-doctor
A health-diagnosis workflow for servers, computers, virtual machines, and containers on Linux, macOS, or Windows.
ai-discover
Parallel discovery of performance hotspots (perf track) and failure surfaces (bug track) for the auto-improvement loop. Fans out one subagent per hot-path area or failure surface; each returns ONE concrete, behavior-preserving fix candidate (perf) or a reproducing test plus fix (bug). Discovery only — no code changes…
debug
Debug issues in the MCP Gateway Registry using first-principles thinking. Invoke when something is broken, timing out, returning errors, or behaving unexpectedly. Forces structured root-cause analysis before any code change is proposed.