Emasoft/ai-maestro-plugin

최근 기록된 소스 활동: 2026년 10월 1일 16:00

SkillsMP 카탈로그 업데이트: 2026년 11월 9일 14:22

수집된 skills: 29

GitHub 스타: 1
GitHub 포크: 0

AI 어시스턴트로 설치

이 프롬프트를 복사해 사용 중인 AI 어시스턴트에 입력하세요.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/Emasoft/ai-maestro-plugin.

저장소 README

README 원본 파일: README.md

README 원본 보기

README 전체 펼치기

# AI Maestro Plugin  [![CI](https://github.com/Emasoft/ai-maestro-plugin/actions/workflows/ci.yml/badge.svg)](https://github.com/Emasoft/ai-maestro-plugin/actions/workflows/ci.yml) ![version](https://img.shields.io/badge/version-3.2.6-blue) ![license](https://img.shields.io/badge/license-MIT-green)  The umbrella core plugin for the AI Maestro ecosystem — shared skills, AMP messaging, AID identity, governance, kanban, and the universal PRRD/TRDD/Kanban workflow that every role plugin inherits. **Skills:** 29 | **Commands:** 14 | **Scripts:** 15 Python/shell Last updated: 2026-06-16 See the [main repo][repo] for the wider ecosystem. [repo]: https://github.com/Emasoft/ai-maestro-plugins ## Installation Install from the `Emasoft/ai-maestro-plugins` marketplace inside Claude Code: ```text /plugin marketplace add Emasoft/ai-maestro-plugins /plugin install ai-maestro-plugin ``` **`node` must be on `PATH`.** Every hook this plugin registers — session tracking, the `directory-guard` sandbox check, notifications, compaction handoff — runs `scripts/ai-maestro-hook.cjs` under Node. Hooks fail quietly (non-zero exit inside a short timeout), so without Node the whole hook surface is inert with no visible error: AI Maestro never sees the session, and the directory guard stops enforcing. The PRRD/TRDD/Kanban pillar scripts need Python 3.10+ on `PATH`. The AMP/AID shell scripts need the `curl, jq, openssl, base64` CLI tools. The optional `memgrep` note-recall engine is published by the **ai-maestro-janitor** (which CORE declares as a plugin dependency, so it is present wherever CORE is); recall degrades to plain grep without it. Most messaging features also require a running AI Maestro server on `http://localhost:23000`. ## Usage The plugin loads its skills automatically; invoke a workflow with its slash command or by describing the task: ```text /ama-prrd-get # read a PRRD rule by number (every role) /ama-trdd-write # author a TRDD; /ama-trdd-transition moves columns /ama-kanban-render # render the design/spec board (read-only) /team-governance # team governance and COS management /amp-inbox # check the inter-agent message inbox /amp-send <to> <subj> <msg> # send a message to another agent ``` Per-skill usage and examples live in each skill's `SKILL.md`; the AMP commands are documented in the table below. ## Skills | Skill | Description | |---------------------------------|----------------------------------------------| | `ai-maestro-agents-management` | Agent lifecycle management | | `agent-identity` | AID agent identity (Ed25519) | | `agent-messaging` | AMP inter-agent messaging | | `agent-repo-workflow` | Agent repo/branch/PR + task done/blocked | | `debug-hooks` | Hook debugging utilities | | `mcp-discovery` | MCP server discovery | | `memory-search` | Wiki-memory recall via memgrep | | `network-security` | Network security checks | | `planning` | Task planning (persistent files) | | `ama-prrd-get` | Read a PRRD rule by number (any role) | | `ama-prrd-find` | Search PRRD rules by content / metadata | | `ama-prrd-edit` | Mutate a SILVER PRRD rule (MANAGER-gated) | | `ama-prrd-propose` | Propose a PRRD change (any role; non-binding)| | `ama-trdd-find` | Find TRDD task-design docs (read-only) | | `ama-trdd-write` | Author a new TRDD (tier-aware zone) | | `ama-trdd-update` | Edit an existing TRDD's body / evidence | | `ama-trdd-transition` | Move a TRDD between columns (matrix-enforced)| | `ama-kanban-render` | Render the design/spec board (read-only) | | `ama-proposal-approvals` | Batch approve/refuse/archive (MANAGER-gated) | | `ama-trdd-server` | TRDD dashboard: approve / search TRDDs | | `ama-panel` | Push HTML to the agent's side panel | | `ama-session` | Queue commands / answer pending prompts | | `ama-unblock` | MANAGER/COS resume of blocked sessions | | `ama-continuity` | Own 5h/7d window health + self-resume | | `ama-portfolio` | R28 approval / mandate tokens (mint, verify) | | `ama-settings` | Gated settings.json editor | | `ama-statusline` | Fleet rate-limit windows at zero API cost | | `team-governance` | Team governance and COS management | | `team-kanban` | Team kanban boards and tasks | ## Memory: transcripts vs curated notes (two complementary systems) Two memory surfaces answer different questions: | System | Skill(s) | Corpus | Question it answers | |--------|----------|--------|---------------------| | Conversation memory | `memory-search` (this plugin) | AI Maestro's indexed conversation transcripts | "what did we SAY / discuss / decide?" | | Wiki note memory | `/janitor-memory-{recall,write,update}` (janitor global) | curated, symptom-indexed wiki pages | "what did we LEARN that must not be re-derived?" | The curated-note memory is now the **janitor's GLOBAL wiki-memory system** (`/janitor-memory-recall`, `/janitor-memory-write`, `/janitor-memory-update`, governed by `~/.claude/rules/markdown-memory-recall.md`). This plugin's own note-memory skills were retired in favor of it; `memory-search` (transcript search) stays and names the global skills as its complement. The `memgrep` engine that recall depends on is **owned and published by the [ai-maestro-janitor](https://github.com/Emasoft/ai-maestro-janitor)**, which CORE declares as a plugin dependency — so it is installed wherever CORE is. Recall degrades to plain `grep` without it. CORE used to vendor its own copy and ship rival binaries; ownership was ruled to the janitor in [ai-maestro#106](https://github.com/Emasoft/ai-maestro/issues/106) and CORE's copy was removed. It was a strict subset — no `validate`, `lint`, `new-page`, `add-atom`, `add-lesson` — published under the same binary name and an identical `version` string (0.1.0 on both), so whichever build landed last silently decided whether the machine-wide memory protocol could run. ## AMP Commands (Agent Messaging Protocol) 12 slash commands for inter-agent communication: | Command | Description | |-------------------|--------------------------------------------| | `/amp-init` | Initialize agent identity and messaging | | `/amp-identity` | Quick identity check for context recovery | | `/amp-status` | Show messaging status and registrations | | `/amp-inbox` | Check message inbox | | `/amp-read` | Read a specific message | | `/amp-send` | Send a message to another agent | | `/amp-reply` | Reply to a message | | `/amp-delete` | Delete a message | | `/amp-register` | Register with an external AMP provider | | `/amp-fetch` | Fetch messages from external providers | | `/amp-download` | Download attachments from a message | | `/amp-statusline` | Install AMP status line for Claude Code | ## AMP Scripts 14 shell scripts installed to PATH for messaging operations: - `amp-init.sh`, `amp-identity.sh`, `amp-status.sh` - `amp-inbox.sh`, `amp-read.sh`, `amp-send.sh` - `amp-reply.sh`, `amp-delete.sh`, `amp-register.sh` - `amp-fetch.sh`, `amp-download.sh`, `amp-statusline.sh` - `amp-helper.sh`, `amp-security.sh` ## AID Scripts (Agent Identity) 5 shell scripts for agent identity management: | Script | Description | |-------------------|------------------------------------------| | `aid-init.sh` | Initialize Ed25519 agent identity | | `aid-register.sh` | Register identity with a provider | | `aid-status.sh` | Show identity status | | `aid-token.sh` | Generate/exchange identity tokens | | `aid-helper.sh` | Shared helper functions for AID scripts | ## Versioning and removal policy Other plugins cite this plugin's skills, commands, and scripts by name in prose, so a removal is a breaking change even when no code imports anything (ai-maestro#118: two removals shipped as MINOR bumps and dependents' citations dangled silently). The policy, enforced by `tests/test_surface_removal_policy.py` at publish time: - **Removing or renaming any agent-facing surface** (a skill, a command, an AMP/AID script, a CLI verb) is a **MAJOR** version bump. No exceptions for "nobody uses it". - **Every removal ships a one-release TOMBSTONE first**: the surface's file stays for at least one release as a stub whose body starts with `TOMBSTONE` and names the successor (or states there is none). The next MAJOR may then drop the stub. - The CHANGELOG entry for the MAJOR names every removed surface. A tombstone only helps dependents who look; the complementary citation-resolver gate (each dependent verifying that skill names cited in its live surfaces still resolve against the provider's released tag) is a generic plugin-quality check that belongs fleet-wide — see the discussion in ai-maestro#118. ## Requirements External tools the plugin's shell scripts call: - `curl(1)` — HTTP requests to AMP providers and AI Maestro - `jq` — JSON parsing in shell scripts - `openssl` — Ed25519 keypair generation for AID - `base64` — message attachment encoding ## Storage AMP/AID state is written under `~/.agent-messaging/`: ```text ~/.agent-messaging/ ├── config.json # local agent config ├── keys/ # Ed25519 keypair (private + public) ├── messages/ │ ├── inbox/ # received messages │ └── sent/ # sent messages ├── registrations/ # external provider registrations └── attachments/ # downloaded attachments ``` ## License MIT

저장소 README를 불러오는 중…

이 저장소의 skills

직업 카테고리 6개 · 83% 분류됨

skill

직업 분류

설명

업데이트

ama-trdd-transition

프로젝트 관리 전문가

Move a TRDD between kanban columns (3-pillars 3.0.0 pipeline) enforcing the role-gated transition matrix — mover, frontmatter side-effects, AMP broadcast. Use when advancing a task. Trigger with /ama-trdd-transition, 'move TRDD X to testing', 'dispatch this…

원문 언어: 영어

2026-10-01

team-kanban

미분류

Manage team kanban boards and tasks. Use when creating, moving, filtering, getting, or editing tasks. Trigger with 'what's on the board', 'move this task to dev', 'create a kanban task', 'show the blocked tasks'. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-09-29

agent-messaging

미분류

Use when sending or receiving inter-agent messages via AMP. Trigger with /amp-send, /amp-inbox, /amp-read. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-09-28

ama-trdd-server

미분류

Search, read, verify, and server-mediated create/mutate TRDDs via the aimaestro-trdd.sh CLI — the AI Maestro dashboard's authorization-aware view of the design/ corpus. Write verbs (create / edit / approve / refuse / promote / archive) are STRICT, gated by…

원문 언어: 영어

2026-09-25

ama-proposal-approvals

미분류

Batch-approve, refuse, or archive TRDD proposals — move them between the four design/ zones (proposals/ to tasks/ on approve; to refused/ on refuse; to archived/ on complete/cancel/supersede). GATED: only a MANAGER (or solo USER) may approve/refuse/archive;…

원문 언어: 영어

2026-09-25

team-governance

컴퓨터·정보 시스템 관리자

Use when managing teams or governance titles. Trigger with /team-governance. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-09-25

ama-trdd-write

미분류

Author a new TRDD task-design document under design/ with correct v2 frontmatter — uuid + timestamps, the right zone (tasks/ for a Tier-0 task written directly as 'planned'; proposals/ for a Tier-1/2/3 task needing approval), and a self-contained body.…

원문 언어: 영어

2026-09-25

debug-hooks

소프트웨어 개발자

Debug Claude Code hooks (PreToolUse, PostToolUse, etc.). Use when hooks aren't firing or produce wrong output. Trigger with 'my hook isn't firing', 'why did the PreToolUse hook not run', 'the hook output is wrong'. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-08-28

ai-maestro-agents-management

기타 컴퓨터 관련 직업

Manage AI agent lifecycle via CLI. Use when creating, listing, deleting, configuring, or looking up an agent's consolidated config (launch args, title, role-plugin, teams, GitHub repo, Docker, pending tasks, AID public key). Trigger with 'create an agent',…

원문 언어: 영어

2026-08-15

ama-session

기타 컴퓨터 관련 직업

Drive an agent's own terminal and read its own activity state via the frozen aimaestro-session.sh CLI — send or enqueue an allowlisted slash command (works even while busy/hibernated), type raw text into a live pane, read the 5-state activity, and read/answer…

원문 언어: 영어

2026-08-08

ama-unblock

기타 컴퓨터 관련 직업

MANAGER/CHIEF-OF-STAFF workflow (governance R42.8) for a stuck agent session — blocked on an AskUserQuestion, a permission prompt, or a rate-limit/API-error banner. Diagnose with block-state plus read-prompt, then resume it by answering its own pending…

원문 언어: 영어

2026-08-08

ama-prrd-propose

소프트웨어 개발자

File a PRRD-change proposal for MANAGER (or USER) review without mutating the PRRD. This is the path EVERY non-MANAGER role uses to request a rule change, and the path for ANY GOLDEN-rule change (which is USER-only). Proposals are non-binding — they queue for…

원문 언어: 영어

2026-08-07

ama-trdd-find

소프트웨어 개발자

Find TRDD task-design documents under design/ — by 8-char id (TRDD-9a8aba94), column, assignee, blocked-by, cited PRRD rule, a SQL-ish --where filter, or free-text; also validate a TRDD's frontmatter. Read-only. Use when locating a task spec before working on…

원문 언어: 영어

2026-08-07

ama-trdd-update

소프트웨어 개발자

Edit an existing TRDD's body or non-column frontmatter as work proceeds — bump updated:, append implementation-commits/ci-runs/test results, write the STATE block, add review notes or post-mortems, set labels/priority/severity. Only the TRDD's current-owner…

원문 언어: 영어

2026-08-07

ama-kanban-render

소프트웨어 개발자

Render the project's design/spec kanban board as a READ-ONLY view over the TRDD pile (the TRDD files under design/ ARE the board) — the full board, just the RED (blocked) column with priority ranking, one column, grouped by assignee, or as JSON; also…

원문 언어: 영어

2026-08-07

ama-prrd-edit

소프트웨어 개발자

Mutate a project's PRRD rules — add/revise/delete a SILVER rule, or promote/demote between SILVER and GOLDEN. GATED: direct mutation is MANAGER-only (enforced by the script's AID_AUTH check); GOLDEN edits and promote/demote are USER-only; a non-MANAGER agent…

원문 언어: 영어

2026-08-07

ama-prrd-find

소프트웨어 개발자

Search a project's PRRD rules by content or metadata — find rules whose text matches a keyword, list all GOLDEN or SILVER rules, find which rules a TRDD cites, find unused rules, or get rule-count stats. Use when an agent asks 'which rule covers X', 'are…

원문 언어: 영어

2026-08-07

ama-prrd-get

소프트웨어 개발자

Read a PRRD rule by number from a project's design/requirements/PRRD.md. Use when an agent needs the exact text of a project rule it must comply with, wants to cite a rule (PRRD G64.134), list all rules, or bootstrap an empty PRRD. Reading is allowed for…

원문 언어: 영어

2026-08-07

ama-panel

소프트웨어 개발자

Drive an agent's HTML dashboard side panel via the frozen aimaestro-panel.sh CLI — push HTML or a live URL, open/close/refresh the panel, check how many dashboard clients are connected, and drain feedback events the panel's HTML posted back. Use when an agent…

원문 언어: 영어

2026-08-06

ama-continuity

기타 컴퓨터 관련 직업

Read this host's rate-limit and cache health, and keep the agent's own session alive, via the frozen aimaestro-continuity.sh CLI — the 5h/7d window percentages, cache TTL, and a next_action recommendation, plus idempotent self-resume and self-restart. Use…

원문 언어: 영어

2026-08-04

agent-identity

정보 보안 분석가

Use when managing agent identity — Ed25519 keys, proof of possession, OAuth token exchange. Trigger with 'create agent identity', 'register with auth server', 'get API token', 'check my registrations'. Loaded by ai-maestro-plugin.

원문 언어: 영어

2026-08-02

agent-repo-workflow

소프트웨어 개발자

Use when an agent needs to clone or create repos, branch, open PRs, list project repos/team members, or report task done/blocked to its orchestrator via AMP. Trigger with 'clone the project repo', 'open a PR for this task', 'list my team's repos', 'report…

원문 언어: 영어

2026-08-02

ama-settings

소프트웨어 개발자

Read and mutate a settings.json / settings.local.json safely via the frozen aimaestro-settings.sh CLI — get, set, delete, and atomic multi-op edit, gated to files living directly inside a .claude directory. Use instead of hand-editing settings JSON, and…

원문 언어: 영어

2026-08-02

ama-statusline

기타 컴퓨터 관련 직업

Read the fleet's 5h/7d rate-limit windows at ZERO API cost via the frozen aimaestro-statusline.sh CLI — the windows arrive free in Claude Code's statusline payload, so get/list hand them back without spending a call on the usage endpoint. Use when checking…

원문 언어: 영어

2026-08-02

ama-portfolio

기타 컴퓨터 관련 직업

Mint, list, verify and revoke R28 approval and mandate tokens via the frozen aimaestro-portfolio.sh CLI — the machine-checkable answer to 'is this agent actually authorized to do this, for THIS card'. Use when granting a one-shot approval or a standing…

원문 언어: 영어

2026-08-02

memory-search

소프트웨어 개발자

Recall prior knowledge from the markdown wiki-memory corpus via memgrep and the janitor's global recall skills. Use before acting on a recurring problem, a design decision, or a repeated alert. Trigger with /memory-search. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-08-02

mcp-discovery

기타 컴퓨터 관련 직업

Discover MCP server tools, resources, and prompts without installing plugins. Use when exploring MCP servers or debugging connections. Trigger with 'what tools does this MCP server expose', 'explore an MCP server without installing it', 'my MCP connection…

원문 언어: 영어

2026-08-02

planning

소프트웨어 개발자

Persistent markdown planning files for complex tasks. Use when starting multi-step tasks needing 5+ tool calls. Trigger with 'make a plan for this', 'track this multi-step task in a plan file', 'resume the plan we started'. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-08-02

network-security

네트워크·컴퓨터 시스템 관리자

Network security model, Tailscale VPN access, IP filtering, and remote device connectivity. Use when configuring remote access or debugging connectivity. Loaded by ai-maestro-plugin

원문 언어: 영어

2026-06-24

수집된 skill 29개 중 29개를 표시합니다.