project
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project manifests and user-provided decision data, creating an attack surface where malicious instructions embedded in project files could influence the agent's behavior. * Ingestion points: Reads project manifest files (PROJECT.md, STUDIO.md), register files (DOCUMENTS.csv, TASKS.csv), and user-provided inputs for facts and decisions as described in SKILL.md and context-resolution.md. * Boundary markers: The instructions in SKILL.md specify a rigorous mutation sequence and context resolution logic to find valid project roots. However, there are no explicit delimiters or specific 'ignore instructions' warnings used when interpolating file content into the agent's context. * Capability inventory: The skill utilizes file read/write operations and shell commands to manage project records as defined in the host-contract.json and SKILL.md. * Sanitization: The skill emphasizes verification via readback in SKILL.md to ensure data integrity after writing, but does not document specific input sanitization or filtering logic for data read from files before it is processed by the model.
Audit Metadata