session-governance-audit
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell and git commands to verify claims made in session transcripts. These include
git log,git show,test -f, andgrep. These are used for validation purposes within the scope of the auditing task. - [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data including session transcripts and task plans. It lacks specific instructions for the agent to distinguish between data content and instructions, creating a surface for injection attacks.
- Ingestion points: The skill reads
session transcript(JSONL/markdown) andoriginating task listdocuments. - Boundary markers: None. There are no explicit instructions to use delimiters or ignore embedded instructions within the processed transcripts.
- Capability inventory: The agent can read the filesystem, execute git commands, create files, and commit changes to the repository.
- Sanitization: None. The skill does not define methods for escaping or validating the content of the transcripts before processing.
Audit Metadata