hook-management
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides commands such as
/hook:disablewhich allow for the intentional bypassing of safety guidelines and agent behavior constraints, specifically targeting session-scoped security hooks like 'file-guard'. - [INDIRECT_PROMPT_INJECTION]: The skill implements a hook architecture that ingests and processes untrusted data from multiple stages of the agentic loop, creating a vulnerability surface.
- Ingestion points: Data is ingested through 'PostToolUse' (tool outputs) and 'UserPromptSubmit' (user prompts) hooks as defined in SKILL.md.
- Boundary markers: The skill documentation does not specify any delimiters or instructions for the agent to ignore embedded commands within the ingested data.
- Capability inventory: The skill allows the use of 'Bash', 'Read', 'Grep', and 'Glob', and references hooks that perform 'test-project' and 'lint-project' which likely execute shell commands.
- Sanitization: There is no evidence of sanitization, filtering, or validation performed on the external content before processing.
Audit Metadata