anti-ai-slop
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external proposal drafts, Terms of Reference (ToR), and research sources. This untrusted data represents a surface for indirect prompt injection if the source materials contain instructions intended to divert the agent from its assigned guardrails.
- Ingestion points: According to the 'Capability Contract' and 'Inputs' table in SKILL.md, the skill requires read access to 'planned proposal content', 'ToR', 'proposer profile', and 'verified sources'.
- Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the ingested draft or source text.
- Capability inventory: The skill allows for 'Search or network verification' and 'editing' within the drafting task, providing a functional path for injected instructions to influence external actions or modify content.
- Sanitization: No explicit sanitization or filtering of external content is specified before interpolation into the agent's reasoning process.
Audit Metadata