claude-coordinator

Plugin claude-coordinator in dennisonbertram/claude-coordinator. Commands read from the repo's manifest — 1 month ago.

Subagents (14)

• briefer

  Context reader and situational analyst. Reads files and returns structured briefings for the coordinator. Fast, thorough, and interpretive.

• coordinator

  Pure-delegation coordinator. Agent-only control plane — delegates ALL I/O to specialized subagents (briefer, planner, worker variants, reviewer, scribe, testers, validators, learning-extractor).

• intent-validator

  Validates that completed work actually satisfies the user's original intent — not just the task instructions, but what the user truly wanted. Runs foreground so it can ask the user directly.

• learning-extractor

  Analyzes session work — task outputs, reviewer findings, intent-validator output, AND sub-agent JSONL transcripts — to surface learnings. Captures both successful patterns and process struggles (retries, dead ends, scope drift, confusion) so future sessions improve.

• planner

  Architecture and task planning agent. Analyzes requirements and codebase to produce task breakdowns with dependencies, file boundaries, and contracts.

• reviewer

  Read-only code reviewer that identifies bugs, regressions, missing tests, and security/concurrency hazards.

• scribe

  Lightweight state writer for .coord/ and docs/ files. Cheap, fast, precise. Handles all file writes the coordinator needs.

• system-tester

  Integration and system-level tester. Runs full test suites, checks regression coverage, validates component integration, and identifies untested code paths.

• ui-tester

  Visual quality tester. Inspects the running UI for layout issues, broken elements, overlapping components, responsiveness, and modern design standards. Uses browser automation.

• ux-tester

  Usability and experience tester. Evaluates whether the app works in a way that makes sense to a human — navigation logic, information architecture, progressive disclosure, and overall intuitiveness.

• worker-investigation

  Read-only research worker. Performs root-cause analysis, codebase exploration, dependency mapping, and feasibility checks. Returns structured findings. Makes no code changes and produces no commits.

• worker-refactor

  Behavior-preserving refactor worker. No new tests required, but the existing relevant test suite MUST pass before and after the refactor. Single commit per task with before/after test evidence.

• worker-test

  Adds tests to existing untested code. Not TDD (code already exists), but tests must be meaningful — they would fail under mutation. Two-commit audit trail per task (tests + verification).

• worker

  Strict TDD implementation worker for `feature` and `bugfix` tasks. Produces an auditable commit trail (red → green → regression) so the test-first practice is provable from git history.

Similar plugins