The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform reconnaissance on the local codebase. It executes commands such as ls, cat, and grep in SKILL.md to detect the presence of package managers, framework-specific directories, and agent configuration files (e.g., .claude/, AGENTS.md). These commands are used for their intended purpose of project discovery.
[COMMAND_EXECUTION]: The reference guide references/skills-reference.md documents the use of dynamic context injection (!command syntax) for agent platforms. It provides examples of using git log, git status, and GitHub CLI (gh) to inject live data into agent prompts. This is documented for educational purposes and reflects standard platform capabilities.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted data from the codebase without explicit sanitization or strict boundary markers.
Ingestion points: Project manifest files (e.g., package.json, pyproject.toml, Cargo.toml) and directory structure results from ls and cat operations.
Boundary markers: Absent. The skill does not instruct the agent to use specific delimiters or to disregard instructions embedded within the analyzed file contents.
Capability inventory: The skill has access to Read, Glob, Grep, and Bash tools, providing a path for potential command execution if the agent is successfully manipulated by malicious content in a processed file.
Sanitization: Absent. Data read from the project is interpolated into the agent's context to determine recommendations without filtering.

agent-automation-recommender