Atelier

Plugin atelier from marketplace atelier in overnightworks/marketplace. Commands read from the repo's manifest — yesterday.

Skills (14)

• aco

  Operate aco (agent claims, the work board, landing) in any repository that coordinates with it. Use before `aco start`, `aco claim`, `aco land`, writing a pull request body, or editing an item body; when briefing agents or releasing a claim; and whenever aco refuses something.

• agent-documentation

  Audit, improve, or seed AI-agent guidance files. Use when changing or reviewing AGENTS.md, CLAUDE.md, agent skills, custom agents, or hooks; when entering a repository whose AGENTS.md/CLAUDE.md is missing, empty, or missing whole policy areas; checking generic versus project-specific ownership; find

• audit

  Audit a codebase through read-only lenses, chosen by goal. Use when asked to clean up, slim down or shrink code, find dead or unnecessary code or the wrong architecture; to check architecture, module boundaries, ownership or duplicate deciders (who decides X); to review security, performance, leaks

• board-hygiene

  Check an open aco/GitHub work board against reality and give every open item a verdict (keep with a reason, close with a reopen trigger, merge, re-cut, body stale, label idea, idea) with evidence, judging every open item against a named current goal. Finds landed-but-open items, twins (also across r

• design-interface

  The workshop's visual language for every user-facing surface — apply whenever building or changing UI (pages, cards, buttons, badges, labels, dialogs, states, motion). Encodes the operator's standard as an enforceable contract: understandable without reading, one glance per question, state by shape

• docs-cleanup

  Clean and audit documentation trees. Use when asked to organize docs, move plans out of documentation into issues, fix documentation drift, reconcile contradictions, repair dead links, or review docs for stale claims. For AGENTS.md, CLAUDE.md, skills, custom agents, and hooks, use agent-documentatio

• explorative-testing

  Exploratively test a running application like a curious user. Use when asked whether a feature actually works beyond its unit tests, to probe a live app for behavior or UX defects, to explore an app against its documented intent, or to turn exploration findings into work items or bug reports.

• legal-texts

  Write, adapt or check one app's own part of the privacy notice (Datenschutzerklärung) and of the terms of use (Nutzungsbedingungen). Reads the current composed pages and shared sections, inventories what the app's code really processes, reads the current official text of each law it checks (GDPR Art

• plan-review

  Review implementation plans before coding. Use when asked to vet a plan for clarity, scope, acceptance criteria, dependencies, contradictions, or implementability.

• providers

  Start a Codex or Grok agent from Bash as a head. Use before running `codex exec`, `codex` or `grok` yourself outside the lane workflow, and whenever a sandbox, a linked worktree or its gitdir blocks a commit, fetch or merge.

• see

  Inspect rendered visual output by capturing and viewing screenshots of it yourself. Use after changing a web page, HTML document, UI, canvas, or visual asset, or when asked whether something is readable, laid out correctly, or visually rough. Launching the app is the run skill's job; building a char

• sonar-distributor

  Turn a repository's SonarCloud and CodeQL findings into exactly ONE distributor issue with the findings classified and sliced — never a fix-by-fix march and never an item flood. Use when asked what Sonar or CodeQL reports for a repository, when a repository is newly onboarded to SonarCloud or CodeQL

• visualize

  Produce accurate diagrams-as-code for documentation. Use when a document needs an architecture overview, dependency graph, sequence, state, flow, or data-model diagram, or when asked to visualize a system, process, or structure for human readers.

• write-docs

  Author or restructure documentation calibrated to its audience, altitude, and document type (README, reference, runbook, architecture overview, vision; the Diátaxis modes guide the writing inside them). Use when asked to write documentation for humans or for AI agents, create an architecture overvie

Subagents (12)

• capability-critic

  Generic read-only critic for changes to skills, custom agents, and agent workflows.

• capability-improver

  Generic worker for improving reusable skills, custom agents, and agent workflow guidance.

• code-auditor

  Generic read-only lens auditor for the audit skill: architecture, tests, code hygiene, runtime risks, and security.

• code-explorer

  Read-only agent for finding the real code paths, owners, and evidence before a change is made.

• code-fixer

  Generic implementation agent for resolving concrete review findings.

• code-reviewer

  Generic read-only reviewer for judging a bounded diff.

• code-worker

  Generic implementation agent for the code phase of a bounded change.

• documentation-curator

  Project documentation worker for updating docs after a bounded repository change.

• explorative-tester

  Generic agent that exploratively tests a landed change through its real entry point, writing only to its own scratch path.

• phase-planner

  Read-only planner for decomposing a long-running objective into verified phases, each one coherent change.

• plan-reviewer

  Read-only reviewer for critiquing implementation plans before work starts.

• refine

  Read-only cataloguer for proposing the expectations an operator may have omitted.

MCP servers (1)

• aco

  runs: aco mcp

Similar plugins