committing-code
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from the repository's staged changes to generate commit messages.\n
- Ingestion points: Processes staged code changes through
git diff --stagedin SKILL.md.\n - Boundary markers: The instructions do not define delimiters or specific warnings to ignore instructions embedded within the staged code.\n
- Capability inventory: The skill has the capability to write to a local file (
COMMIT_MESSAGE.md) and execute shell commands (git commit -F).\n - Sanitization: Staged content is not sanitized or escaped before processing to generate the commit body.\n- [COMMAND_EXECUTION]: The skill workflow involves executing local git commands to interact with the repository.\n
- Evidence: Uses
git diff --stagedandgit commit -F COMMIT_MESSAGE.mdas specified in the workflow section of SKILL.md.\n - Context: Command usage is limited to the skill's primary purpose and uses predefined file paths and arguments.
Audit Metadata