agent-atelier

Plugin agent-atelier from marketplace ether-plugins in ether-moon/ether-plugins. Commands read from the repo's manifest — 1 month ago.

Skills (4)

• execute

  Run the development loop end-to-end. Auto-runs plan cycle if no valid plan_approval exists; otherwise drives IMPLEMENT → VALIDATE → REVIEW_SYNTHESIS → AUTOFIX → DONE. Use when starting or resuming work after planning. Triggers on 'execute', 'run', 'go', 'start', 'continue', 'pick up where we left of

• monitors

  [INTERNAL — invoked by orchestrator/cron, not for direct user use] Background monitor lifecycle — spawn continuous monitors, poll for events, stop monitors, or check health. Triggers on 'spawn monitors', 'check monitors', 'stop monitors', 'monitor status', 'spawn ci monitor', 'poll events', 'respawn

• plan

  Run a planning cycle (DISCOVER → SPEC_DRAFT → SPEC_HARDEN → BUILD_PLAN) with ping-pong clarifying questions and a final approval gate. Use when starting a new feature, when spec or work items need rework, or when the user wants explicit plan review before execution. Triggers on 'plan', 'planning', '

• status

  Show the orchestration dashboard — current loop mode, plan state (active plan cycle, plan_approval, recent ping-pong activity), work item summary, active candidate set, candidate queue, open gates, and next action. Use when checking progress, understanding what's happening, reviewing the backlog, re

Subagents (7)

• architect

  Bridge between Behavior Spec and executable implementation — decomposes spec into vertical-slice work items with file ownership, sets complexity, populates verify, sequences Builders, and resolves merge conflicts. Spawned by the Orchestrator as an always-on core teammate; routes state writes through

• builder

  Ephemeral full-stack implementer — spawned per work item, implements frontend/backend/API/DB to satisfy the WI's Behavior, runs unit and integration tests, produces atomic commits. Conditional specialist; shut down after WI completion or requeue. Spawn with mode: plan for complex WIs.

• pm

  Owner of product truth — writes and revises the Behavior Spec, classifies feedback (bug | spec_gap | ux_polish | product_level_change), maintains assumptions and decision logs. Spawned by the Orchestrator as an always-on core teammate; never implements code or runs tests.

• qa-reviewer

  QA Reviewer — interprets VRM evidence bundles for spec compliance and defect detection. Identifies functional defects, distinguishes spec gaps from implementation bugs, and assigns severity. Conditional specialist spawned during REVIEW_SYNTHESIS. Read-only by design — never runs validation tooling,

• state-manager

  Sole writer for .agent-atelier/** workflow state — receives structured update requests, validates against the current revision, and commits atomically through the state-commit script. Routes control-plane mutations only (status transitions, mode changes, candidate lifecycle, promotion, completion).

• ux-reviewer

  Pragmatic UX Reviewer — interprets VRM evidence for usability, accessibility (WCAG), and intuitiveness. Identifies missing affordances, confusing flows, accessibility violations with specific WCAG citations. Conditional specialist spawned during REVIEW_SYNTHESIS. Read-only by design — web validation

• vrm

  Validation Runtime Manager — runs full test suites, Playwright scenarios, and accessibility checks against candidate branches; assembles evidence bundles at .agent-atelier/validation/<run-id>/. Conditional specialist spawned per validation run. Operates strictly from the build-vrm-prompt input — nev

Hooks (7)

• UserPromptSubmit

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-prompt.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-pre-tool-use.sh

• Stop

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-stop.sh

• SubagentStop

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-stop.sh

• TaskCompleted

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-task-completed.sh

• TaskCreated

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-task-created.sh

• TeammateIdle

  ${CLAUDE_PLUGIN_ROOT}/hooks/on-teammate-idle.sh