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/Liberty91LTD/cti-skillsnpx agentmods add skills/liberty91ltd/cti-skills/quality-controlWrote 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/liberty91ltd/cti-skills/quality-control)<a href="https://agentmods.dev/skills/liberty91ltd/cti-skills/quality-control"><img src="https://agentmods.dev/badge/skills/liberty91ltd/cti-skills/quality-control/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/liberty91ltd/cti-skills/quality-control"><img src="https://agentmods.dev/badge/skills/liberty91ltd/cti-skills/quality-control.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 | $0.00056 | $0.02652 |
| Opus 5.5 | $0.00022 | $0.01061 |
| Sonnet 5.5 | $0.00011 | $0.00530 |
| Haiku 4.5 | $0.00006 | $0.00265 |
Grade A, and why
quality-control 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 1mo 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 — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Quality Control Standards
Review Process
- All products must be reviewed before moving to "published" status
- Reviews are performed by the quality-reviewer agent
- Products with CRITICAL issues are returned for revision
- Products with 3+ MAJOR issues are returned for revision
- MINOR issues are noted but don't block publication
- Evidence validation runs before the checklist. A product that fails it is returned without being scored.
Evidence Validation
This is the enforcement point for /claim-extraction and /quality-of-information-check output, and for any product built on it (/ach, /threat-assessment, /writing-assessments). The schema is in skills/quality-of-information-check/references/evidence-item-schema.md. The caps are rules R1 to R13 in skills/quality-of-information-check/references/grading-rubric.md.
Two passes, in this order.
Pass 1: Script check
python3 skills/quality-of-information-check/scripts/validate_evidence.py <qoi.json> [--source-text <file>]
- Exit 0: valid. Go to Pass 2.
- Exit 1: violations, listed on stdout as JSON. Each one is a CRITICAL issue. Return the output for revision with the list attached. Do not fix grades or anchors yourself.
- Pass
--source-textwhenever the source document text is available, so anchors can be checked against it. - If the script cannot be run, say so in the review and do every check in the table below by hand. Do not report the script check as passed.
Pass 2: Reviewer check
The script covers what can be checked mechanically. The reviewer then reads the claim table against the source and confirms each rule below. Any failure is CRITICAL and the output is rejected.
| # | Reject when | How to check |
|---|---|---|
| V1 | An anchor is missing | Every item has anchor.sentence and anchor.location, non-empty |
| V2 | Anchor text is not present verbatim in the source | Find anchor.sentence in the source text, character for character. A paraphrase fails |
| V3 | A claim contains two assertions | One subject, one assertion. A compound "and" joining two facts that could be graded differently fails |
| V4 | claim_type is not one of the five values |
observation, attribution, assessment, actor_claim, victim_disclosure. press_originated is a flag, not a type |
| V5 | A grade exceeds a cap | Rules R1 to R13. See the table below |
| V6 | Corroboration is asserted without basis | independent_primaries of 2 or more needs a basis other than unchecked and a sources entry for each additional primary. The same applies to any "corroborated by" wording in the rationale or the product text |
| V7 | provenance_basis is absent |
Present on every item and in the output header, with one of platform-resolved, script-resolved, model-judged |
| V8 | Claim count is outside five to twelve for a primary | Count claims per primary document. Claims from attributed quotes in an article are additional and not counted. Allowed outside the range only when the user asked for it, and the output says so |
| V9 | A grade is not two words | grading.access_level and grading.claim_support each hold one of the rubric's values. The old fields source_reliability and information_credibility, a letter, a number or a code all 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.
- 1mo ago Changed · +65 lines · +35 tokens per session caeebf34e4fe
- 2mo ago First seen · 84 lines · 21 tokens per session scan A 999c615aa82b
quality-control is a skill published in the GitHub repository Liberty91LTD/cti-skills (36 stars, last pushed 4d ago), licensed MIT. It adds 56 tokens to every session and 2,652 once invoked, about $0.0002 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 2026-09-03.
Other skills, from other repositories
analyzing-campaign-attribution-evidence
Systematically evaluate cyber-campaign evidence to attribute an operation to a threat actor, using the Diamond Model and Analysis of Competing Hypotheses (ACH) to weigh infrastructure overlaps, TTP consistency, malware code similarity, and timing/language artifacts into confidence-weighted attribution assessments. Use…
analyzing-threat-actor-ttps-with-mitre-attack
MITRE ATT&CK is a globally-accessible knowledge base of adversary tactics, techniques, and procedures (TTPs) based on real-world observations. This skill covers systematically mapping threat actor beh.
analyzing-threat-landscape-with-misp
Analyze the threat landscape using MISP (Malware Information Sharing Platform) by querying event statistics, attribute distributions, threat actor galaxy clusters, and tag trends over time. Uses PyMISP to pull event data, compute IOC type breakdowns, identify top threat actors and malware families, and generate threat…
analyzing-campaign-attribution-evidence
Use when campaign attribution analysis involves systematically evaluating evidence to determine which threat actor or group is responsible for a cyber operation. This skill covers collecting and weighting attr.
prompt-injection-auditor
Security audit of LLM system prompts, agent instruction files (SKILL.md, AGENTS.md, CLAUDE.md), and agent configurations against prompt injection attacks. Use when the user wants to (1) audit or harden a system prompt or agent instructions against prompt injection, (2) review an agent skill or system prompt for…
analyzing-campaign-attribution-evidence
Campaign attribution analysis involves systematically evaluating evidence to determine which threat actor or group is responsible for a cyber operation. This skill covers collecting and weighting attr.