The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing external documentation files likely authored by third parties. * Ingestion points: The skill reads CHANGELOG.md, MIGRATION.md, and UPGRADING.md files from project dependencies and local directories. * Boundary markers: The skill relies on simple markdown headers to separate ingested content, which does not provide robust protection against embedded instructions. * Capability inventory: The skill can execute git commands, write and delete files in project memory, and dispatch high-capability tools like 'build' for codebase refactoring. * Sanitization: There is no evidence of input validation or instruction filtering performed on documentation content before it is interpolated into agent prompts.
[COMMAND_EXECUTION]: The skill performs automated shell and filesystem operations for project state management and maintenance. * It executes git commands (branch and rev-parse) to initialize tracking markers and capture environment metadata. * It manages its own lifecycle by writing status updates and deleting stale scratch directories older than 48 hours. * It orchestrates complex code changes by delegating execution tasks to the 'build' skill.

migrate