What problem does it solve? Managing configuration for multiple AI agents (Hermes, Claude Code, Codex, Cursor, OpenCode, Gemini) across computers leads to duplicated skills, drifting variants, and the risk of committing secrets. This Skill provides a policy and workflow for building a small, auditable repository that syncs agent settings without credentials, caches, or runtime state. ## Core Features & Use Cases - Shared skill deduplication: Canonicalize skills by declared frontmatter name, remove agent-specific duplicates, and expose shared skills via symlinks. - Secret-safe configs: Replace MCP tokens and credentials with ${ENV_VAR} references, exclude OAuth stores and sessions, and scan for leaked secrets. - Snapshot and install workflow: Deterministic snapshot scripts, dry-run-by-default installers, and cross-machine round-trip procedures with conflict policy. - Use Case: You set up a new laptop and want your Claude Code and Codex skills, MCP servers, and settings restored. Clone the portability repo, run the installer in dry-run mode, then apply skills and reauthenticate OAuth servers. ## Quick Start Ask the agent to snapshot the current machine's agent skills and configs into a secret-safe portability repository and verify there are no duplicate shared skill names.
