block-issue
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exposes an unvalidated data ingestion surface by passing untrusted user input directly to a shell script.
- Ingestion points: User input is ingested via the
$ARGUMENTSvariable inSKILL.md. - Boundary markers: No explicit boundary markers or isolation techniques are employed to separate user arguments from the command execution flow.
- Capability inventory: The skill possesses command execution capabilities, invoking the local script
"${CLAUDE_PLUGIN_ROOT}/scripts/larch.sh"using the Bash tool. - Sanitization: The input handling only removes the
--run-idflag, leaving other arguments unvalidated before processing.
Audit Metadata