rabbit-inspect
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local commands using
uv runto trigger theglimpsetool located in the project'stools/glimpsedirectory for screenshot capture and matrix generation. - [COMMAND_EXECUTION]: The skill uses the
gwtool to interact with GitHub for creating issues based on the audit results. - [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface as it ingests and processes content from external websites.
- Ingestion points: Reads and analyzes content/screenshots from target URLs provided by the user in
SKILL.md(e.g., Phase 2: LOOK and Phase 3: WANDER). - Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat website content as untrusted data or to ignore embedded instructions within the pages being audited.
- Capability inventory: The skill has the capability to write to external systems by creating GitHub issues via the
gw gh issue createcommand in Phase 5: BURROW. - Sanitization: There is no evidence of sanitization or escaping of the content retrieved from the target URL before it is used to generate reports or GitHub issue bodies.
Audit Metadata