gh-address-comments
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/fetch_comments.pyexecutes the GitHub CLI (gh) via thesubprocess.runfunction to retrieve pull request metadata, reviews, and comments. This is standard and legitimate use of local tooling for PR management. - [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data by fetching GitHub PR comments and review bodies which are then used to guide the agent's actions.
- Ingestion points: The
scripts/fetch_comments.pyscript retrieves thebodyfield of comments and reviews from the GitHub GraphQL API. - Boundary markers: The skill does not currently employ specific delimiters or "ignore embedded instructions" warnings when presenting comment content to the agent.
- Capability inventory: While
fetch_comments.pyonly reads data, the agent is subsequently instructed inSKILL.mdto "Apply fixes," which grants it the capability to modify the local filesystem and source code based on the content of those comments. - Sanitization: There is no evidence of sanitization or filtering of the fetched comment text before it is processed by the agent.
Audit Metadata