Borrowing it
Nothing to install: this file belongs to dfirtnt/Huntable-CTI-Studio. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dfirtnt/Huntable-CTI-Studio/main/.claude/skills/eval-fixture-audit/SKILL.mdgit clone --depth 1 https://github.com/dfirtnt/Huntable-CTI-StudioWrote 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/dfirtnt/huntable-cti-studio/eval-fixture-audit)<a href="https://agentmods.dev/skills/dfirtnt/huntable-cti-studio/eval-fixture-audit"><img src="https://agentmods.dev/badge/skills/dfirtnt/huntable-cti-studio/eval-fixture-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/dfirtnt/huntable-cti-studio/eval-fixture-audit"><img src="https://agentmods.dev/badge/skills/dfirtnt/huntable-cti-studio/eval-fixture-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 | $0.00209 | $0.02950 |
| Opus 5 | $0.00105 | $0.01475 |
| Sonnet 5 | $0.00042 | $0.00590 |
| Haiku 4.5 | $0.00021 | $0.00295 |
Grade A, and why
eval-fixture-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 12d 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 — 228 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Eval Fixture Audit (one extractor per run)
Independently re-derive the eval expected count (Eval1) and ground-truth item list (Eval2) for every eval article of a single extractor agent, strictly from that agent's spec — then compare against all recorded sinks, adjudicate divergences with the operator, and apply only approved corrections.
Run ONE agent per session. The recorded counts are anchors: if you read them before doing your own extraction, you will rationalize toward them (two prior model audits diverged on the same articles for exactly this reason). The discipline is extract-first, compare-second.
Execution requirements — interactive only, fail-closed
This skill's ONLY safety mechanism is a human operator answering the ⛔ STOP gates (Step 1 rubric confirmation, Step 6 sink-by-sink write approval). That mechanism does not exist in a headless / autonomous / cloud run — so the skill must fail closed, never free-wheel:
- Do NOT run this audit without an interactive operator. If you were dispatched autonomously and no human will answer mid-run, STOP NOW and report that this audit requires an interactive operator. Do not extract, write, commit, or push.
- At every ⛔ gate, no operator response = HALT. Do not write any sink, do not commit, do not push, do not self-approve to keep moving. A run that stops at a gate and reports "blocked: need operator" is a SUCCESS; a run that routes around a gate is a FAILURE even if its edits are correct.
- Never act as the operator. "The operator confirms / directs / approves" always means the human. You propose; only the operator disposes.
- Never push. Pushing is always a separate, explicit operator instruction — never
an autonomous step, even on a
claude/*session branch. - A spot-fix is not an audit. The run is not complete until Steps 3–5 produced a blind per-article extraction for EVERY article and a divergence table across ALL four authoritative sinks (xlsx, yaml, articles.json, ground_truth.json). The DB is derived run output, not a fixture sink. Spotting one bad item by reading ground_truth.json is the anchored shortcut this skill exists to prevent — not a passing run.
What ships with it
2 files 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.
- 12d ago First seen · 228 lines · 209 tokens per session scan A 52485a062a26
eval-fixture-audit is a skill published in the GitHub repository dfirtnt/Huntable-CTI-Studio (11 stars, last pushed 4d ago), licensed MIT. It adds 209 tokens to every session and 2,950 once invoked, about $0.0010 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 skills, from other repositories
configuring-windows-event-logging-for-detection
Configures Windows Event Logging with advanced audit policies to generate high-fidelity security events for threat detection and forensic investigation. Use when enabling audit policies for logon events, process creation, privilege use, and object access to feed SIEM detection rules. Activates for requests involving…
sast-dast-coordinator
../../../appsec-devsecops/sast-dast-coordinator/SKILL.md.
dast-workflow
Dynamic Application Security Testing workflow — OWASP ZAP automation (baseline/full/API scans), Burp Suite Professional playbooks, Burp Collaborator for out-of-band detection, auth-state orchestration, and CI integration with scope-safe active scanning.
redamon-testing
How RedAmon tests actually run and how to author them: the per-file Docker gate, the unit/integration/live tiers, and the failure modes that make a green run a lie. Trigger: editing any test.py, .test.ts(x) or tests/.sh; a test that is red, skipped or xfailed; a request to "run the tests", "make it green" or check…
exploiting-jwt-algorithm-confusion-attack
Exploits JWT algorithm confusion vulnerabilities where the server's token verification library accepts the algorithm specified in the JWT header rather than enforcing a fixed algorithm. The tester manipulates the alg header to switch from RS256 to HS256 (using the RSA public key as the HMAC secret), sets alg to none…
performing-soap-web-service-security-testing
Perform security testing of SOAP web services by analyzing WSDL definitions and testing for XML injection, XXE, WS-Security bypass, and SOAPAction spoofing.