diagnose
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data to reproduce bugs, which creates a surface for indirect prompt injection attacks where malicious instructions could be embedded in the data being debugged.
- Ingestion points: The agent is instructed to "replay a captured trace" (Phase 1.5), save network requests/payloads/event logs (Phase 1.5), and process external artifacts such as HAR files, log dumps, and core dumps (Phase 1 fallback).
- Boundary markers: The instructions do not define clear delimiters or warnings to treat the content of these debug artifacts as untrusted data that should not be followed as instructions.
- Capability inventory: The skill has significant local capabilities, including executing shell commands (CLI, curl), running headless browsers (Playwright/Puppeteer), performing
git bisectoperations, writing new test files, and committing changes to a repository. - Sanitization: There are no instructions for sanitizing or filtering the content of ingested traces or logs before they are processed by the agent.
- [COMMAND_EXECUTION]: The skill explicitly directs the agent to execute a variety of shell-based tools and scripts, including custom human-in-the-loop (HITL) bash scripts, test runners, and CLI tools, which is necessary for its debugging purpose but involves the execution of dynamically generated or modified scripts.
Audit Metadata