preen-deferred-fixes

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and acts upon untrusted data from GitHub issues and PR comments.
  • Ingestion points: The skill reads the body of GitHub issues and PR comments using ./scripts/agents/tooling/agentTool.ts actions like getIssue and findDeferredWork (SKILL.md).
  • Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or isolate potentially malicious commands embedded within the fetched PR comments.
  • Capability inventory: The skill has significant capabilities, including executing local scripts (agentTool.ts), running shell commands (pnpm), and performing code modifications via git (/commit-and-push, git push).
  • Sanitization: The analysis of the instructions reveals no sanitization or validation logic to filter out executable instructions found within the PR review data.
  • [COMMAND_EXECUTION]: The skill executes local scripts and system tools to perform its tasks.
  • Evidence: It invokes ./scripts/agents/tooling/agentTool.ts to interact with GitHub APIs.
  • Evidence: It executes pnpm lint, pnpm typecheck, and pnpm test for code validation.
  • Evidence: It uses git commands for version control operations. These commands are consistent with the skill's intended purpose of managing code fixes.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 10:12 PM