The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to generate and run local shell scripts, test suites, and network commands (e.g., curl) to create a deterministic reproduction of the reported bug.\n- [COMMAND_EXECUTION]: The skill includes a pre-defined bash script scripts/hitl-loop.template.sh which the agent executes to interact with the user, provide instructions, and capture command-line responses.\n- [PROMPT_INJECTION]: The skill processes untrusted data from user-provided logs and bug descriptions, creating an attack surface for indirect prompt injection. This is noted as a risk inherent to the debugging task, and the skill includes phases for cleaning up temporary instrumentation.

diagnosing-bugs