The Agent Skills Directory

[SAFE]: The skill provides a structured framework for continuous improvement through response templates and quality checkpoints. It does not perform inherently dangerous operations or execute scripts.- [NO_CODE]: The skill consists entirely of markdown instructions and a metadata report. No executables, shell scripts, or library dependencies are included, which limits the technical attack surface.- [PROMPT_INJECTION]: The skill establishes workflows for processing user feedback and interaction history, which represents an indirect prompt injection surface. Ingestion points: User feedback strings and response logs described in SKILL.md. Boundary markers: The templates utilize markdown code blocks for data organization but do not specify unique delimiters for untrusted input. Capability inventory: Limited to platform-native tools (Read, Write, Grep, Glob). Sanitization: The skill instructions recommend validation-first coding and security-review checkpoints in the provided frameworks.

creating-feedback-loops