The Agent Skills Directory

[SAFE]: The skill defines a logical workflow for debugging that relies on observation and evidence-based analysis rather than arbitrary code execution.
[DATA_EXFILTRATION]: The instructions involve inspecting environment variables and git history to diagnose issues. These actions are standard for the debugging context and do not involve sending data to external domains.
[PROMPT_INJECTION]: No patterns attempting to bypass safety filters or override system instructions were found. The skill possesses an indirect prompt injection surface as it processes external logs and source code:
Ingestion points: Error logs, stack traces, and source code files (SKILL.md)
Boundary markers: Absent
Capability inventory: Shell execution for tests, file system modifications, and git operations
Sanitization: Absent
[REMOTE_CODE_EXECUTION]: No remote scripts or unauthorized package installations are performed. The skill recommends running local tests and reproduction commands, which is expected behavior for a debugging tool.

nv-debug