--- type: Guide title: "GUIDE-444 — Agent Skills Standard (SKILL.md): The 444-SUITE Capability Spec {v4.33.7-r2}" description: "GUIDE-444 = the Agent Skills Standard — the canonical SKILL.md capability specification for ♾️ WeOwnNet/WeOwn.Buzz BuzzAgents. Within the 444 SUITE (AGENTS.md = identity · SKILL.md = capability · MCP = transport), GUIDE-444 defines the SKILL.md payload format adopted ecosystem-wide from the open AgentSkills.io standard (Anthropic-originated, cross-vendor), per the 8/8 MetaCouncil consensus. Authored by FedArchBuzz as deliverable of GTM_2026-W33_7108 track (Gitea push), REF 7109." tags: [guide-444, 444-suite, agents-md, skill-md, agentskills, skillmd, responsibleagenticai, build-the-pipeline, metacouncil, fedarchbuzz, weownnet, s004, okf, bp-075, w33] resource: "https://git.weown.tools/WeOwnAI/s004_fedarch/raw/branch/main/_GUIDES_/GUIDE-444.md" sources: - id: consensus-8of8 resource: "mem/agentskills-skillmd-consensus.md (parent REF GTM_2026-W33_6135, tracker 6076)" author: "🏛 MetaCouncil (8/8 unanimous)" last_modified: "2026-08-16" note: "8/8 NO OBJECTIONS → EMBRACE AgentSkills.io; 10 core SKILL.md practices; avg score 82 (Sage 97 top)." - id: deepresearch-verdict resource: "OUTBOX/CONTEXT_BROADCAST_SKILL_STANDARDS_v4.33.6-r1.md (REF 6135, tool-verified 2026-08-15)" author: "AI:FedArchBuzz 🐝" last_modified: "2026-08-15" note: "SKILL.md singular = standard (cross-vendor verified); SKILLS.md plural = legacy aggregate/index only." - id: prj444-r4 resource: "https://git.weown.tools/WeOwnNet/s004/raw/branch/main/_PROJECTS_/PRJ-444.md (v4.33.7-r4)" last_modified: "2026-08-16" note: "444 SUITE charter; §2.5 GUIDE-444 sync (designation LOCKED @GTM fff4908d; tracker 7008)." - id: agent-skills-io resource: "https://agentskills.io (HTTP 200, 267,638 B, verified 2026-08-15)" note: "Open standard; org by Anthropic Oct 2025, open Dec 2025; 37-46+ clients; SKILL.md Required + optional scripts/references/assets." status: draft generated: by: "AI:FedArchBuzz 🐝 (DeepSeek V4 Flash 0731)" at: "2026-08-16T23:00:00Z" verified: [] --- # 📋 GUIDE-444 — Agent Skills Standard (SKILL.md) ## 🏆 #GoldStandard — v4.33.7-r2 (WeOwnVer) · W33 D7 Sun · #LLMmodel: DeepSeek V4 Flash 0731 (OpenRouter) > **GUIDE-444 is the Agent Skills Standard for ♾️ WeOwnNet 🌐 — the canonical SKILL.md capability specification.** It closes the 444-SUITE loop: **AGENTS.md = who you are · SKILL.md = what you can DO · MCP = how you transport.** Every BuzzAgent ships a conformant `SKILL.md`, and GUIDE-444 is the spec that governs it. --- ## 1 · PURPOSE & SCOPE | Field | Value | |:------|:------| | **Document** | GUIDE-444 | | **Version** | v4.33.7-r2 ✅ (WeOwnVer) | | **Folder (SOT)** | `_GUIDES_/` 📋 — **Gitea ORG WeOwnAI/s004_fedarch** (`git.weown.tools/WeOwnAI/s004_fedarch`, GUIDE-015/415.1 precedent) | | **Suite** | 444 SUITE = AGENTS.md (identity) + **SKILL.md (capability)** + MCP (transport) | | **Category** | 📋 GUIDE — Agent Skills Standard | | **Lifecycle** | 🟡 **DRAFT — staged for push** (R-011 ⬜ @GTM:ADMIN) | | **Designation** | LOCKED as **GUIDE-444** per @GTM event `fff4908d` (2026-08-16) · tracker `GTM_2026-W33_7008` · L-468 GUIDE-4xX (S004 = 4xy) | | **Author** | AI:FedArchBuzz 🐝 (DeepSeek V4 Flash 0731) | | **REF** | GTM_2026-W33_7109 (this deliverable) · tracker GTM_2026-W33_7108 | | **Standard** | Google OKF v0.2 (Open Knowledge Format) + BP-075 verifiable footer | --- ## 2 · THE ONE-LINE STANDARD > **`SKILL.md` (singular, uppercase) inside `skills//` is THE capability payload.** `SKILLS.md` (plural) is allowed ONLY as a root-level index/README — never the payload. Verified cross-vendor (tool-first 2026-08-15, agentskills.io HTTP 200 · Anthropic docs 200 · Claude Code docs 94,693 B): this is the open AgentSkills.io standard, originally developed by Anthropic, released open Dec 2025, adopted by 37-46+ clients (Claude Code, Codex/ChatGPT, Copilot, Cursor, Gemini CLI, VS Code). **Our own Buzz harness already conforms** (`.agents/skills/buzz-cli/SKILL.md`, on-disk PoP 11,014 B). --- ## 3 · 10 CORE PRACTICES (MetaCouncil 8/8 verdict, deduped) | # | Practice | |:-:|:---------| | 1 | `SKILL.md` singular = payload · `SKILLS.md` plural = root index only | | 2 | Frontmatter `name` + `description` min; `version` mandatory (WeOwnNet) | | 3 | Kebab-case folders; **folder name = command** | | 4 | Self-contained folder + optional `scripts/` `references/` `assets/` | | 5 | Treat `SKILL.md` as executable code (Aegis P0: review scripts, sandbox, provenance) | | 6 | Progressive disclosure: Discovery → Activation → Execution | | 7 | Independent semver per skill; Drift Gate additive-only | | 8 | Complementary: AGENTS.md=who · SKILL.md=how · MCP=transport | | 9 | Portability test on 2+ platforms before MAIT-template lock | | 10 | EN body; bilingual descriptions + `references/zh/` for F1Visa/APAC (Sage) | > **Council scores (1-100):** Sage 97 · DRP-P08 96 · Aegis 95 · DRP-P05 93 · DeepPro 73 · Sprout 70 · MiMo 68 · Surge 65 · **Avg 82**. Sage = most superior tool-first usage (@GTM observation confirmed). Source: mem/agentskills-skillmd-consensus (parent REF 6135, tracker 6076). --- ## 4 · REQUIRED FRONTMATTER (MINIMUM) ```yaml --- name: # folder name = command description: version: # mandatory (WeOwnNet) --- ``` Additional recommended (per OKF + suite): `author`, `license`, `dependencies`, `tags`. Unknown keys MUST NOT be rejected (OKF v0.2 §4.4). --- ## 5 · SKILL FOLDER LAYOUT ```text skills// ├── SKILL.md # 🔴 THE payload (Required) ├── scripts/ # optional executable helpers ├── references/ # optional deep docs (/zh/ for bilingual) └── assets/ # optional static resources ``` --- ## 6 · PROGRESSIVE DISCLOSURE (CORE PATTERN) 1. **Discovery** — directory + frontmatter make the skill findable. 2. **Activation** — agent reads `SKILL.md` body on invoke (name/description match). 3. **Execution** — scripts/references load on demand; body is a pointer, not a dump. This keeps context lean and matches BP-404 + our load-on-demand discipline. --- ## 7 · SECURITY (Aegis P0 — not optional) - Treat `SKILL.md` as executable code: **review scripts before use, sandbox execution, verify provenance.** - Never auto-run unknown skill scripts on a shared/hosted agent surface. - `scripts/` = reviewed, versioned, signed-in-practice where feasible. --- ## 8 · COMPLEMENTARITY WITH THE 444 SUITE | Doc | Role | Says | |:----|:-----|:-----| | AGENTS.md | Who you are | identity, governance, cage, non-negotiables | | **SKILL.md** | **What you can DO** | **capability spec ← THIS (GUIDE-444)** | | MCP | How you transport | tool/skill transport layer | --- ## 9 · HOW TO ADOPT (STEPS) 1. Author each skill as `skills//SKILL.md` with min frontmatter. 2. Keep payload singular `SKILL.md`; use `SKILLS.md` only as root index. 3. Independent semver per skill; Drift Gate additive-only (rN+1 ≥ rN). 4. Test portability across 2+ skills-compatible clients before locking into a MAIT template. 5. EN body; add `references/zh/` for bilingual/APAC (Sage). --- ## 10 · PROVENANCE & GOVERNANCE - **Status:** 🟡 DRAFT — staged for push (REF 7109 · tracker 7108). R-011 ⬜ NOT assumed (#127). - **Push** = @GTM:ADMIN / iMAIT (FedArchBuzz stages, never pushes). Target: `_GUIDES_/GUIDE-444.md` on **Gitea ORG WeOwnAI/s004_fedarch** (git.weown.tools, authored @GTM directive 2026-08-16 22:53:36Z; GUIDE-015/415.1 precedent — confirmed present). - **Verify** = live raw-200 + BP-075 `--verify` exit 0 after push. - **Full context:** PRJ-444 v4.33.7-r4 (charter + §2.5 sync) · CONTEXT_BROADCAST_SKILL_STANDARDS (verdict) · mem/agentskills-skillmd-consensus (verdict). --- ## ✅ REVISION TABLE (rN → rN+1) | Version | Change | Drift Gate | |:--------|:-------|:-----------| | v4.33.7-r1 | Initial spec authored (this artifact, GH target) | — baseline | | **v4.33.7-r2** | **SOT re-anchored to Gitea ORG WeOwnAI/s004_fedarch/_GUIDES_/** (was GH CCCbotNet; @GTM 22:53:36Z) | ✅ PASS | --- **REF GTM_2026-W33_7109 · NEXT 7110 · masterCCC GTM_2026-W32_7029 · s004d080 (Sun 16 Aug 2026) · #GUIDE444 #AgentSkills #SKILLmd #444Suite #ResponsibleAgenticAI #BuildThePIPELINE #FedArchBuzz #MetaCouncil #OKFv02 #BP075 #FlowsBros** ♾️ WeOwnNet 🌐 [DUNA] Ecosystem ● 🏡 Real Estate and 🤝 cooperative ownership for everyone ● An 🤗 inclusive community, *by 👥 invitation only*. ### ✅ BP-075 CANONICAL HASH GENERATED [@GTM:ADMIN generated @ 2026-08-16 16:56 MDT | #HITL by @yonks 21:06 MDT] Content-SHA256: 3aa6e25fbb1cab898619912f35a2b33ab60c1857734b098998cab4c623db6971 FEDARCH-CANARY: 3aa6e25f CHARACTERS: 8538 WORDS: 1183 LINES: 167