The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests content from repository files (such as AGENTS.md, CLAUDE.md, and CONTRIBUTING.md) to populate a 'Review Rubric' in the generated context file. This information is intended to guide the behavior of other agents in the suite. Ingestion points: Repository documentation and metadata files (references/01-collect-facts.md). Boundary markers: Extracted content is placed within defined Markdown sections and tables in the output file. Capability inventory: The skill performs filesystem writes, git metadata queries, and optional package installations. Sanitization: Data is parsed and formatted into structured Markdown tables; the skill asks for user confirmation before final writes.
[EXTERNAL_DOWNLOADS]: The skill offers an optional installation of the 'fallow' package from the npm registry to support dependency graph analysis. This action is transparent and requires explicit user consent on a permissions screen.
[COMMAND_EXECUTION]: The skill executes various system commands (git churn measurement) and Node.js scripts to perform repository discovery. It also composes package manager commands for dependency installation based on the detected environment. All such actions are scoped to the setup process and subject to user review.

lodestar-setup