The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run various local commands, including test suites, git bisect run, and a provided human-in-the-loop script (scripts/hitl-loop.template.sh).- [DATA_EXFILTRATION]: The workflow requires the agent to handle potentially sensitive diagnostic data such as HAR files and core dumps provided by the user. This is documented as a standard part of the debugging process.- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external files and user-provided input without explicit sanitization. 1. Ingestion points: User input in scripts/hitl-loop.template.sh and external artifacts in Phase 1. 2. Boundary markers: None present. 3. Capability inventory: Command execution, file modification, and network requests via curl. 4. Sanitization: None present.

diagnosing-bugs