codebase-tour
Pass
Audited by Gen Agent Trust Hub on May 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface identified.
- Ingestion points: The skill (specifically in the 'Survey the repo' section) instructs the agent to read arbitrary files from the codebase, including
README.md,AGENTS.md, package manifests, and source code. - Boundary markers: There are no boundary markers or 'ignore embedded instructions' directives specified to isolate the content of analyzed files from the agent's control logic.
- Capability inventory: The skill has permissions to read files across the repository and write documentation files (e.g.,
SYSTEM_OVERVIEW.md) to the local disk. - Sanitization: Content from analyzed files is incorporated into the agent's context without sanitization or validation, allowing potentially malicious instructions in a codebase to influence the agent's behavior.
Audit Metadata