본문으로 건너뛰기
cccskills
무료GitHub에 공개

tdmcp-submission

Orchestrates the tdmcp Anthropic Connectors Directory (Desktop Extension / MCPB) submission prep — designs the package, writes the privacy page, migrates .dxt→.mcpb, drafts all form answers, and QAs the result. Use for ANY work on the tdmcp directory/marketplace submission, including re-runs after a rejection ("update the submission", "redo the form answers", "re-migrate the bundle", "fix the privacy page", "the directory rejected us"). Drives a 4-agent pipeline (architect → docs-author ∥ bundle-engineer → QA).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(1)

  • SKILL.md4.9 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

tdmcp submission orchestrator

Coordinates four agents to prepare (and re-prepare) tdmcp's submission to the Anthropic Connectors Directory via the Desktop Extension (MCPB) path. Defines who works when; each agent's skill defines how.

Execution mode — sub-agents (this environment has no TeamCreate)

TeamCreate/SendMessage are not available here, so run agents as sub-agents with the Agent tool. For every agent:

  • subagent_type: "general-purpose", model: "opus".
  • Prompt MUST tell it to read its definition file in .codex/agents/*.toml and its required skill(s) first, then do the task. Use this exact mapping: architect → .codex/agents/submission-architect.toml, docs-author → .codex/agents/docs-author.toml, bundle-engineer → .codex/agents/bundle-engineer.toml, and submission-qa → .codex/agents/submission-qa.toml. The definition file is the source of truth (reusable next session); the prompt just points at it + passes the task and the relevant _workspace/ paths.
  • Parallelize independent agents by sending multiple Agent calls in one message (or run_in_background: true).

Data passing — file-based via _workspace/

Create _workspace/ at the repo root for inter-agent artifacts:

  • 00_submission-spec.md — architect → everyone
  • 01_migration-notes.md — bundle-engineer → QA/human
  • 02_form-answers.md — docs-author → QA/human (paste-ready form draft)
  • 03_qa-report.md — QA → human

Final deliverables (privacy page, edited repo files, tdmcp.mcpb) land in the repo proper. _workspace/ is preserved for audit / re-runs.

Phase 0: context check (initial vs re-run)

  1. If _workspace/ is absent → initial run: full pipeline below.
  2. If _workspace/ exists and the user asks to fix/update one part → partial re-run: re-invoke only the affected agent(s), pass the prior artifacts + the new feedback (e.g. a directory rejection reason). E.g. "privacy page rejected" → docs-author only, then QA.
  3. If _workspace/ exists and the user gives fresh inputs → move it to _workspace_prev/ and start a new initial run.

Pipeline (initial run)

  1. architect (solo, first). Produces 00_submission-spec.md. Everything downstream depends on it — do not parallelize anything before it finishes.
  2. docs-author ∥ bundle-engineer (parallel — disjoint file sets):
    • docs-author owns: privacy page(s), config.ts nav additions, 02_form-answers.md. (Prose lane.)
    • bundle-engineer owns: dxt/manifest.json, scripts/build-dxt.mjs, package.json scripts, scripts/setup.mjs, .dxt→.mcpb text sweep in existing docs, .github/ release workflow, 01_migration-notes.md. (Code lane.)
    • They never edit the same file — that's why they can run together safely.
  3. submission-qa (solo, last). Runs builds + checks, writes 03_qa-report.md. If it reports blocking issues, loop back to the owning agent (partial re-run), then re-run QA on the affected checks.

Error handling

  • An agent failing once → retry once with the failure context. Still failing → proceed and record the gap in the final summary; never silently drop it.
  • Never bypass a failing gate (--no-verify, deleting a check). A red QA gate is the signal to fix, not to suppress.
  • Conflicting facts (e.g. spec says manifest_version X, packer wants Y) → trust the live tool/repo, and have the architect update the spec rather than forcing the stale value.

After the run

Summarize to the user (in Portuguese): what changed in the repo, the contents of 02_form-answers.md and the QA verdict, and the explicit Needs human before submit list (support email, privacy contact, cutting a new release so the .mcpb asset exists, and the actual form click — which only the user can do). Then offer to address any QA nits or feedback.

Test scenarios

  • Happy path: no _workspace/ → architect spec → privacy page + form draft + .mcpb migration → npm run docs:build and build:mcpb pass → QA all-PASS except human-input items → summary lists those items.
  • Rejection re-run: user says "directory rejected: privacy policy incomplete" → Phase 0 detects existing _workspace/ → re-invoke docs-author with that feedback → QA re-checks the privacy gate only → updated summary.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Use when manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hybridlabor-api/aos62026년 10월 16일 업데이트

Use when a hybrid memory system that provides persistent, searchable knowledge management for AI agents (Architecture, Patterns, Decisions).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hybridlabor-api/aos62026년 10월 16일 업데이트

Reference for how BDB structures autonomous software engineering work — the seven-node dispatcher graph (Architect, TechLead, UI/UX, Engineering, Media/EventTech, Reviewer, Shipping) that /startcycle-graph actually runs. Use when you need the high-level lifecycle framing without inventing your own process.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hybridlabor-api/aos62026년 10월 16일 업데이트

Tools are how AI agents interact with the world. A well-designed tool is the difference between an agent that works and one that hallucinates, fails silently, or costs 10x more tokens than necessary. This skill covers tool design from schema to error handling.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hybridlabor-api/aos62026년 10월 16일 업데이트

Harness patterns for coding agents — memory, permissions, context engineering, delegation, skills, hooks, bootstrap.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hybridlabor-api/aos62026년 10월 16일 업데이트

agenttrail: live map of a multi-agent build in the browser: which plan component is being worked on, by which agent or harness, what is done and what is stuck. Use when a multi-agent pipeline starts (/startcycle, /startcycle-graph, /teamwork-preview) or after a plan-canvas approve, when the user asks to see what the agents are doing ("live map", "build board", "who is running now"), or by yourself whenever a multi-agent run is under way — no user prompt needed.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hybridlabor-api/aos62026년 10월 16일 업데이트

hybridlabor-api의 스킬 모두 보기

이 스킬의 문제 신고