The Agent Skills Directory

Standard Command Execution: The skill uses Bash to perform codebase analysis (e.g., ls, cat, grep). This is a necessary and standard part of the skill's functionality to identify project types and dependencies.
Indirect Prompt Injection Surface: The skill ingests data from external sources within the codebase, such as package.json, project structures, and dependency lists. While this represents a surface area for indirect prompt injection, it is managed by the agent's internal guardrails and is considered a low-severity consideration.
Ingestion points: SKILL.md (via cat package.json, ls -la commands).
Boundary markers: Not explicitly present in the analysis phase.
Capability inventory: Uses Read, Glob, Grep, and Bash for file system inspection.
Sanitization: Standard output filtering is used (e.g., head -50, grep).
Educational Dynamic Context Injection: The documentation provides templates for creating new skills that use the !command`` syntax to inject live data (e.g., git log, gh pr diff). These are educational examples provided to the user rather than active execution patterns within this specific skill's runtime.

claude-automation-recommender