commit
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_CONTEXT_INJECTION]: The
SKILL.mdfile uses the!command`` syntax to rungit branch,git status,git diff, andgit logwhen the skill is loaded. These commands are standard for development environments and are used here to provide the agent with local context. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local repository environment which could be manipulated.
- Ingestion points: Repository information, including diff stats and commit logs, is loaded into the agent's context in
SKILL.mdvia dynamic context injection. - Boundary markers: There are no explicit delimiters or warnings to the model to ignore instructions that might be embedded in the repository data.
- Capability inventory: The agent is instructed to use commands such as
git branch,git switch,git commit, andgit resetbased on the processed data. - Sanitization: The skill does not implement sanitization or validation of the content retrieved from the git repository before it is processed by the agent.
Audit Metadata