The Agent Skills Directory

[SAFE]: The skill consists of markdown-based procedural guidelines and YAML metadata. It does not include executable scripts, automated package installations, or remote code execution patterns.\n- [SAFE]: The skill explicitly forbids the installation of external providers ("Never installs providers") and requires verification of scope and boundaries before proceeding with implementation.\n- [SAFE]: External references to documentation for Claude Code, OpenCode, and the Pi coding agent are provided for informational purposes regarding agent runtimes and do not involve untrusted downloads or execution.\n- [SAFE]: Potential indirect prompt injection surfaces, such as GitHub issues, are addressed through a multi-stage review process involving triage, lead-agent oversight, and mandatory independent read-only reviews of all code changes.

deliver-spec