developer-audience-context
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill builds a developer audience context by analyzing external, potentially untrusted sources from the project environment.
- Ingestion points: Instructions in
SKILL.mddirect the agent to analyze the codebase, specifically targetingREADME.md,GitHub Issues,existing blog posts, andlanding pages(e.g.,index.html). These sources may contain instructions or content provided by external third parties. - Boundary markers: The skill instructions lack explicit delimitation or strict boundary markers (such as XML tags or unique delimiters) when analyzing these files, which increases the likelihood of the agent accidentally executing instructions found within the data.
- Capability inventory: The skill performs file system read operations across the project and file system write operations to the
.agents/developer-audience-context.mdfile (as defined inSKILL.md). - Sanitization: The
SKILL.mdfile contains a 'Research evidence and privacy' section that provides behavioral guidelines (e.g., minimizing private data and avoiding invented quotes), but no technical filtering or escaping of the ingested content is present.
Audit Metadata