The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell commands for repository setup, specifically creating symlinks (ln -sf) and verifying file contents (cat). It also incorporates the gh (GitHub CLI) tool in generated sub-skills for managing pull request reviews, which is a standard developer tool.
[DYNAMIC_EXECUTION]: The skill dynamically generates additional agent skill files (SKILL.md) by adapting provided reference templates to the target repository's specific environment. This script generation is a core feature for configuring local agent workflows and uses predefined structures.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external repository data (such as package.json, CI configurations, and existing documentation) to inform the generation of project-specific documentation and rules.
Ingestion points: Project configuration files, lockfiles, and existing documentation within the target repository.
Boundary markers: None explicitly implemented during the automated data gathering phase.
Capability inventory: File system write access and basic shell command execution.
Sanitization: The skill employs natural language instructions for the agent to avoid secrets and ensure accuracy rather than programmatic filtering.
[DATA_EXFILTRATION]: While the skill analyzes repository contents, it includes explicit safety constraints forbidding the inclusion of secrets or environment variables in the generated documentation, instead directing agents to reference example files like .env.example.

bootstrap-agents-md