draft-blog-post
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local PowerShell scripts (
upsert-blog-draft.ps1,run-daily-draft.ps1) using the-ExecutionPolicy Bypassparameter. This is used to ensure the skill's internal scripts can run on systems with restrictive execution policies. - [DYNAMIC_EXECUTION]: The
upsert-blog-draft.shshell script executes embedded Python code by piping it directly to the interpreter (`python3 - <<'PY'`). This allows complex logic to be contained within a single shell script.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests git commit history and diffs from external repositories to generate blog content. This introduces a surface where malicious instructions in commit messages could attempt to influence the agent.
- Ingestion points: Commit history and diffs fetched via
gitcommands inSKILL.md. - Boundary markers: Not explicitly defined for commit message processing.
- Capability inventory: File system writes to
docs/blogand git commit operations. - Sanitization: Content is summarized by the AI without pre-processing filters for malicious instructions.
- [PERSISTENCE]: The skill includes
SCHEDULING.md, which provides instructions for users to set up recurring automation using Windows Task Scheduler, cron, and GitHub Actions. This facilitates the long-term, automated operation of the blog drafting process.
Audit Metadata