reports
Pass
Audited by Gen Agent Trust Hub on Mar 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requests access to the
Bash(python3 *)tool, which enables the execution of arbitrary Python code to locate and process files. (SKILL.md) - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by design.
- Ingestion points: Reads content from
report.md,state.json, andmetadata.jsonartifacts generated by previous runs in theruns/directory. (SKILL.md) - Boundary markers: The instructions explicitly require labeling boundaries between artifact content and fresh interpretation, which provides a partial mitigation. (SKILL.md)
- Capability inventory: The skill uses
Read,Glob, andBash(python3 *)to access and process potentially untrusted data from these files. (SKILL.md) - Sanitization: There is no explicit requirement to sanitize or validate the content of retrieved reports before displaying or processing them. (SKILL.md)
Audit Metadata