diagnose
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill describes a workflow that involves executing various command-line utilities for reproduction and testing, such as
go test,pnpm test,curl, andgit bisect. These commands are standard for the skill's primary purpose of diagnosing software bugs. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external, potentially untrusted sources such as user-provided terminal input and captured log artifacts (e.g., HAR files, event logs).
- Ingestion points: User feedback captured by
scripts/hitl-loop.template.shand external debugging artifacts mentioned in Phase 1 (Replay a captured trace). - Boundary markers: The instructions do not define specific delimiters to isolate user-provided error messages or logs from the agent's instruction context.
- Capability inventory: The skill grants the agent capabilities to modify code, execute shell scripts, and perform performance profiling.
- Sanitization: No explicit sanitization or validation of the captured output is performed before it is processed by the agent.
Audit Metadata