The Agent Skills Directory

[SAFE] (SAFE): The skill content is purely instructional and focused on legitimate software engineering best practices. The instructions encourage a scientific approach to debugging, including evidence gathering and hypothesis testing.
[DATA_EXPOSURE] (SAFE): While the skill mentions environment variables and keychain diagnostics (e.g., security list-keychains), it does so in the context of troubleshooting CI/CD and build systems. It demonstrates security-aware techniques like ${IDENTITY:+SET} to verify the presence of secrets without exposing their actual values.
[INDIRECT_PROMPT_INJECTION] (LOW): As a debugging tool, the skill naturally interacts with untrusted data such as logs and error messages. Evidence: 1. Ingestion points: Read, Grep, and Glob tools used on external logs and code files. 2. Boundary markers: No explicit delimiters or instruction-ignore headers for processed data. 3. Capability inventory: Bash, Write, and Edit tools are available. 4. Sanitization: No evidence of sanitization for ingested content. This represents a known surface for indirect prompt injection common to debugging tools but is not a malicious feature.
[COMMAND_EXECUTION] (SAFE): The skill's metadata appropriately limits tool usage to standard development utilities (Bash, Grep, Read, Glob, Edit, Write) required for its primary purpose of identifying and resolving software defects.

systematic-debugging