diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions guide the agent to perform various command-line operations, such as executing CLI tools with fixture inputs, using git for automated bisection (e.g.,
git bisect run), and running shell scripts to verify reproduction steps. - [DYNAMIC_EXECUTION]: The skill involves the generation and execution of dynamic scripts and test cases, including HTTP scripts (curl), headless browser scripts (Playwright/Puppeteer), and custom throwaway harnesses designed to isolate and reproduce software defects at runtime.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data sources, including system logs, captured traces, network payloads, and human-provided input through a human-in-the-loop (HITL) script template (
scripts/hitl-loop.template.sh). This data is used to inform the agent's diagnosis and could potentially influence the agent's logic if the ingested content contains adversarial instructions, though the skill includes human verification checkpoints in the loop.
Audit Metadata