pr-report
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several shell commands, including running a local Python script (
contrib/repo_metrics/weekly_pr_report.py), using the GitHub CLI (gh) to view pull request details, and runningmakeandmkdocsfor documentation building and verification.- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted data from external pull requests, creating an attack surface for indirect prompt injection. This content is processed by sub-agents to generate technical summaries and breaking change reports. - Ingestion points: Pull request metadata and code diffs are fetched and stored in
.work/pr_deep/*.json(SKILL.md). - Boundary markers: The skill uses markdown and JSON for structuring sub-agent tasks but lacks explicit delimiters to isolate untrusted code content from the agent's instructions.
- Capability inventory: The skill has the capability to write files and modify core project configuration files like
docs/mkdocs.yml. - Sanitization: The skill includes a step to remove specific sensitive internal terms for public-facing reports but does not sanitize the raw code diffs or PR descriptions for prompt injection patterns.
Audit Metadata