The Agent Skills Directory

[SAFE]: The skill implements a 'diagnose-first' approach with explicit requirements for human approval before any code modifications. Analysis of the instructions and the accompanying Python script reveals no patterns of unauthorized command execution, data exfiltration, or malicious persistence. The use of local file references (e.g., to shared operating principles) is standard for this environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process arbitrary error logs and tracebacks provided by the user. While this constitutes a surface for indirect prompt injection, the risk is inherent to the task of debugging and is mitigated by the skill's conservative policy.
Ingestion points: The scripts/safe_debug.py script reads user-provided content via the --error-text or --error-file arguments.
Boundary markers: The script wraps excerpts of the untrusted input in Markdown code blocks in its output files, but does not implement explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The skill is capable of writing diagnosis reports to the file system and relies on the agent's ability to execute the provided diagnostic script.
Sanitization: No content sanitization or character filtering is performed on the ingested error text; it is processed for keyword classification and excerpted verbatim.

safe-debug