The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes content from other SKILL.md files and profile configurations, creating a surface for indirect prompt injection.
Ingestion points: Uses cue lint-skill, grep, and head to read data from local files (SKILL.md, profiles/profile.yaml).
Boundary markers: No explicit delimiters or boundary markers are used to isolate content from audited files.
Capability inventory: Performs file system operations (mkdir), file reads, and executes local project tools (cue, bin/cue-learnings).
Sanitization: Content from external files is processed without explicit sanitization before being used in shell commands or prompt interpolation.
[COMMAND_EXECUTION]: Employs standard Unix utilities for file analysis and project management.
Evidence: Uses grep, sed, head, and wc to parse and count lines in project files as part of the linting and auditing process.
Evidence: Utilizes mkdir -p to create directory structures for new skill scaffolding.
[COMMAND_EXECUTION]: Interacts with vendor-provided CLI tools and local scripts to perform advanced auditing tasks.
Evidence: Calls the cue binary for linting skills and ranking skill usage based on session history.
Evidence: Executes bin/cue-learnings to record and search discoveries across sessions, maintaining a local knowledge base in the user's home directory.

skill-reviewer