What problem does it solve? Agent harness configuration (hooks, MCP servers, plugin bundles) is scattered across incompatible per-tool formats, and mistakes like wrong root keys, invented hook event names, or plaintext secrets silently break the setup. This Skill provides the authoring rules and per-tool references to write that config correctly once and deploy it everywhere via APM. ## Core Features & Use Cases - Lifecycle hook authoring: Guidance for PreToolUse, PostToolUse, SessionStart and other events, with matchers, exit-code semantics, timeouts, and cross-tool compatibility across Claude Code, VS Code Copilot, and APM. - MCP server configuration: APM-first declaration in apm.yml under dependencies.mcp, stdio/http/sse transports, ${VAR} secret placeholders, and a full per-tool schema matrix covering the servers vs mcpServers root-key trap. - Plugin bundle packaging: plugin.json manifests, marketplace distribution, apm pack workflows, and consumer discovery-budget limits for bundled skills. - Use Case: You need a hook that blocks writes to protected files in every editor your team uses. The Skill routes you to the hooks reference, gives you a PreToolUse matcher pattern, and tells you to author it once in .apm/hooks/ so APM deploys it to each target. ## Quick Start Ask the agent to add a PreToolUse hook that runs a formatter after every file edit and declare a new MCP server in apm.yml using this Skill's guidance.
