The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands including git, bun, and just to perform repository inventory and analysis. It also provides guidance on using ephemeral runners like npx, uvx, or cargo for read-only validation of changes.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from external repositories, which presents a surface for indirect prompt injection.
Ingestion points: Reads files such as AGENTS.md, CLAUDE.md, and project manifests from target repositories using inventory.ts.
Boundary markers: Instructions mandate mechanical verification (diffs, hashes) and evidence attachment for every finding.
Capability inventory: Accesses file system for reading and writing (authorized by user) and executes local development tools.
Sanitization: Explicitly requires adapting content for the recipient and prohibits the transfer of secrets or project-specific credentials.

repo-harmonization