client-report-from-commits
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
gitbinary via a Python script (scripts/collect_git_changes.py) to gather commit history. This is done usingsubprocess.runwith a list of arguments, which prevents shell injection. The repository path and date arguments are validated or resolved to absolute paths before execution. - [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests untrusted data from git commit logs to generate client-facing reports.
- Ingestion points:
scripts/collect_git_changes.pyreads commit subjects and bodies usinggit logoutput. - Boundary markers: While the script outputs structured JSON, there are no explicit boundary markers or instructions to the agent to treat the commit content as untrusted data in the
SKILL.mdprompt interpolation context. - Capability inventory: The skill allows file system interaction via git commands and execution of the included Python collection script.
- Sanitization: No sanitization or filtering is performed on the commit subjects or bodies before they are presented to the LLM for summarization.
Audit Metadata