The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to facilitate the debugging process. This includes running unit tests, integration tests, curl commands, CLI tools, and git bisect. It also provides a dedicated bash template (scripts/hitl-loop.template.sh) for interactive reproduction steps.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data, such as project files, session logs, and user-provided traces, creating a potential attack surface for indirect prompt injection.\n
Ingestion points: Data enters the agent's context through reading the local codebase (referenced in SKILL.md), searching historical session logs, and processing user-provided error messages and artifacts captured via scripts/hitl-loop.template.sh.\n
Boundary markers: The skill does not define specific delimiters for raw data ingestion in SKILL.md, but it enforces a strict rule to redact secrets and authentication headers before the agent processes or displays them.\n
Capability inventory: The skill utilizes shell command execution in SKILL.md (e.g., bash, npm, pip, curl, git, wsl) to verify hypotheses and run reproduction scripts.\n
Sanitization: The instructions in SKILL.md contain explicit safety directives for the agent to proactively redact credentials and use environment variables to ensure secrets are not exposed in logs or shared artifacts.

diagnosing-bugs