pr-link-issue
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell commands via the GitHub CLI (
gh) to read pull request numbers, view issue comments, extract text, and update pull request descriptions (gh pr view,gh issue view,gh pr edit). - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external collaborative environments, establishing a potential surface for indirect prompt injection.
- Ingestion points: Pulls the existing PR description text via
gh pr viewin Step 4, and pulls external comments from the GitHub issue viagh issue viewin Step 3. - Boundary markers: Absent. There are no formatting delimiters or system guidelines instructing the agent to ignore prompt commands embedded within the retrieved PR body or issue comments.
- Capability inventory: Possesses the capability to write and overwrite GitHub PR descriptions using
gh pr editbased on the processed text. - Sanitization: Issue comments are partially restricted using a regex pattern (
grep -Eioe '[a-z]+-[0-9]+'), but the full text of the existing PR body is ingested into the agent context without filtration.
Audit Metadata