skills/thaitype/chief/chief-grill/Gen Agent Trust Hub

chief-grill

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill maintains session state by creating and reading files within a project-local directory (.chief/_grill/). It provides instructions to sanitize user-supplied topics into kebab-case slugs, which mitigates risks associated with malicious filename construction or path traversal.
  • [PROMPT_INJECTION]: An indirect prompt injection surface is present as the skill records user-provided answers into persistent session logs. These logs are subsequently ingested by the agent during resume operations and passed to a background answer-verifier-agent for cross-referencing.
  • Ingestion points: User-supplied session topics and interview answers (SKILL.md), along with findings returned by the answer-verifier-agent (SKILL.md).
  • Boundary markers: No specific markdown delimiters or instructions to ignore embedded commands are defined for the session log content.
  • Capability inventory: File system read/write access scoped to the .chief/ directory and the ability to spawn subagents via the Agent tool (SKILL.md).
  • Sanitization: Filenames are sanitized via kebab-case derivation; however, the skill does not specify filtering or escaping for the content of recorded answers.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 12:35 AM
Security Audit — agent-trust-hub — chief-grill