deep-fact-check
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill consists exclusively of markdown documentation (README.md, SKILL.md) and an example file. It does not include any scripts, binaries, or automated installation hooks that could execute code on the host system.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data (URLs, articles) for fact-checking.
- Ingestion points: External articles and documents provided via URL or text in the
SKILL.mdworkflow. - Boundary markers: The instructions explicitly demand a structured 'Cognitive Deconstruction' using SIFT principles and 'Triple-Layer Fact Modeling', which acts as a logical barrier between the untrusted input and the agent's reasoning.
- Capability inventory: The skill uses standard agent capabilities like web searching and summarization as defined in the methodology.
- Sanitization: The skill mandates a 'Cognitive Bias Self-Check' and 'Epistemic Status' labeling, which encourages the agent to evaluate content critically rather than obeying embedded instructions.
- [EXTERNAL_DOWNLOADS]: The skill mentions installation via
npx skills add Lionad-Morotar/deep-fact-check-skill. This is a standard package management pattern for this agent ecosystem and targets the author's own repository.
Audit Metadata