The Agent Skills Directory

[SAFE]: The skill is entirely instructional and promotes safe, methodical engineering practices such as hypothesis testing and verification before applying code changes.- [PROMPT_INJECTION]: While the skill instructs the agent to analyze external data such as payloads, logs, and database states which could contain indirect prompt injections, this is an inherent part of the debugging process. The risk is minimized as the skill provides a framework for critical analysis rather than automated execution of untrusted input.
Ingestion points: System payloads, message contents, file data, and database state as mentioned in 'Failure Modes to Avoid' and 'Persistent State' sections.
Boundary markers: Not explicitly defined in the methodology.
Capability inventory: File system write operations for logging findings in ~/.qwen/investigations/ and subprocess execution for running test commands.
Sanitization: Not explicitly mentioned, though the methodology emphasizes 'verifying data collection' and 'identifying decision points'.
[SAFE]: The use of the path ~/.qwen/investigations/ for persistence is limited to a vendor-specific hidden directory and is used for documentation purposes across sessions, which is standard behavior for this agent's debugging workflow.

structured-debugging