Engineering Workflows for Claude Code
@maestria/claude-code is a declarative Claude Code plugin: a
.claude-plugin/plugin.json manifest, a directory of agent files, two skills, and three workflow
commands. The agent files, skills, and workflow commands are generated from the canonical
directives in packages/core/agent-directives/; the plugin manifest is maintained separately in
the package.
Status: Native candidate. The plugin validates cleanly with the official CLI (
claude plugin validate --strict) and follows the current documented plugin contract, but runtime behavior is not yet tested end to end. Upstream Claude Code docs are moving and unpinned - reverify before relying on any claim. Do not treat this package as a production support promise. See the runtime support matrix for evidence.
How It Works
Section titled “How It Works”- Plugin loads - Claude Code discovers the
.claude-plugin/plugin.jsonmanifest and theagents/,skills/, andcommands/directories from the plugin directory - Rules preloaded - every specialist agent preloads the
maestria:global-rulesskill via the agentskillsfrontmatter field - Agents available - 7 specialist agents are available as namespaced
@maestria:<agent>mentions - Orchestrate - the
/maestria:orchestratorskill routes work to specialists;/maestria:fein,/maestria:sonar, and/maestria:blitzselect the pipeline depth - Read-only roles enforced -
adventurer,planner, andreviewerdeny theWriteandEdittools viadisallowedTools(user-authorized; the only runtime enforcement in the package)
Quick Start
Section titled “Quick Start”The plugin is a self-contained directory. Validate it from a checkout, or install the published package persistently through the Maestria CLI:
claude plugin validate ./packages/claude-code --strictclaude --plugin-dir ./packages/claude-codenpx maestria install claude-codeFull Installation Guide →