analyze-issues
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the internal
larch.shscript via the Bash tool using arguments forwarded from the user prompt. This design is susceptible to command injection if malicious shell characters (such as;,&, or|) are included in user-provided flags like--repoor--log-root. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill fetches and reads content from external GitHub issues, comments, and repository logs.
- Boundary markers: The instructions state the agent should not parse or branch based on the data, but the resulting insight report is returned directly to the agent's context.
- Capability inventory: The skill has access to
BashandReadtools to interact with the filesystem and execute internal logic. - Sanitization: There is no explicit requirement or description of sanitization or escaping of the ingested GitHub content before it is included in the generated report.
Audit Metadata