review-axes
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources, including git logs, commit messages, and external issue trackers (GitHub, GitLab).
- Ingestion points: Git commit history (
git log), issue/PR bodies from remote trackers, and local markdown files (SKILL.mdstep 2). - Boundary markers: The instructions explicitly direct the agent to place untrusted data inside fenced blocks and include a warning to ignore any instructions within that data.
- Capability inventory: The skill executes
gitcommands, spawns sub-agents, and writes to local files to update checkboxes and status labels (SKILL.mdstep 1, 4, and 6). - Sanitization: The skill relies on prompt-based isolation (fencing) to separate data from instructions rather than structural sanitization.
- [COMMAND_EXECUTION]: The skill executes local shell commands via git to calculate diffs and verify references.
- Commands used:
git diff,git log,git rev-parse, andgit add(SKILL.mdstep 1). - Context: These commands are used to identify the changes to be reviewed and ensure the provided commit references are valid.
Audit Metadata