code-conventional-commit
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands (
git status,git diff --staged, andgit commit -m) to manage the version control workflow as described in its primary operations. - [EXTERNAL_DOWNLOADS]: The file
workspace/iteration-1/report.htmlreferences an external JavaScript library (xlsx.full.min.js) fromcdn.sheetjs.comto provide spreadsheet rendering capabilities for the evaluation reports. - [PROMPT_INJECTION]: The skill analyzes repository changes via
git diff, which creates a surface for indirect prompt injection where malicious instructions embedded in a staged file could attempt to influence the agent's behavior during commit message generation. - Ingestion points: Untrusted content from the user's workspace is read into the agent's context using
git diff --stagedinSKILL.md. - Boundary markers: The instructions do not specify the use of clear delimiters or instructions to ignore embedded commands within the diff content.
- Capability inventory: The skill has the authority to execute shell commands, specifically
git commit -m, as defined inSKILL.md. - Sanitization: There is no explicit sanitization or filtering of the content retrieved from the git diff before it is processed by the model.
Audit Metadata