content-strategy
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
- [NO_CODE]: This skill is entirely comprised of markdown documentation and configuration files (SKILL.md, evals.json, headless-cms.md). It contains no executable code or scripts.
- [PROMPT_INJECTION]: The skill includes instructions to analyze potentially untrusted external data, presenting a surface for indirect prompt injection. 1. Ingestion points: Ideation sources in SKILL.md, including Reddit/Quora forum data, customer call transcripts, and survey responses. 2. Boundary markers: Absent; there are no specific instructions for the agent to ignore embedded commands within the research data. 3. Capability inventory: None; the skill does not possess tools for file system writing, network communication, or arbitrary command execution. 4. Sanitization: Absent; external data is processed directly as text.
- [SAFE]: The skill's instructions to read local files such as .agents/product-marketing-context.md for business information is a standard practice for establishing task context and does not constitute unauthorized data exposure.
Audit Metadata