grill
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill instructions in Phase 0 direct the agent to "read in full any file $ARGUMENTS names" and to read broad project-related files (code, config, docs). While intended for context gathering, this capability could be used to expose sensitive system or configuration files if they are provided as input.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from files named in
$ARGUMENTSand general project code to drive its interview logic, creating an attack surface for embedded malicious instructions. - Ingestion points: Reading external files specified in user arguments and project-level documentation/code in Phase 0.
- Boundary markers: Not present. The instructions do not define specific delimiters to isolate ingested file content from the agent's core instructions.
- Capability inventory: The skill performs file system reads, sub-agent spawning via
spawn_agent, and file writing totmp/reports/. - Sanitization: Not present. No logic is defined to filter or sanitize the content read from files before it is processed by the agent.
Audit Metadata