本文へ移動
cccskills
無料GitHub で公開

gaia-suite-intake

Comprehensive, reusable playbook for multi-component suite intake under Yggdrasil III. Audits component installability, maps origin candidates to generic capabilities, ingests named components with shared-repo baseline caps, manifests suites via gaia dev fuse, calibrates capstone and component ranks, and closes out via Class S regeneration.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md12.9 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

gaia-suite-intake

End-to-end playbook for ingesting, structuring, and calibrating multi-component skill suites under Yggdrasil III.

A suite is a cohesive collection of specialized skills authored by a contributor or team, anchored by a central capstone skill that coordinates or unifies them (e.g. gsd-build/get-shit-done, leonxlnx/taste-skill, obra/superpowers). Ingesting a suite requires orchestrating multiple interlocking layers: component installability, generic capability mappings, shared repository evidence capping, suite manifestation via gaia dev fuse, branch-aware calibration, and Class S artifact regeneration.


Yggdrasil III Normative Rules & Invariants

Before executing any intake steps, verify understanding of these foundational invariants:

  1. Type Axis Collapse:
    • The starless node type is strictly {basic, fusion}.
    • extra, ultimate, and unique are retired as types — never write them to node JSON.
  2. Derived Branch Taxonomy:
    • Branch (standard / suite / unique) is derived dynamically, never hard-coded.
    • A skill carrying suiteComponents sits on the Suite branch at any rank (1★–6★).
    • Standalone skills without suiteComponents sit on standard (1★–3★) or unique (4★–6★).
  3. Trust Magnitude as Sole Promotion Authority:
    • The legacy per-star Evidence Floor is completely retired.
    • fusion-recipe rows contribute 0 TM (Issue #1600). Structural composition is reported separately via Fusion Score, which gates no rank.
  4. Shared Repository Evidence Cap:
    • Sub-components inheriting repository evidence from a shared parent repo are capped at SUITE_COMPONENT_REPOSITORY_CAP = 50.0 TM per component (META.md §2.1, trustMagnitude.py).
    • Repository adoption alone cannot promote sub-components beyond Grade B (50 TM / 3★) without independent corroboration.
  5. Branch-Aware 4★ Gate (META.md §4 / Issue #1741):
    • 4★ Promotion Requirement: Requires Overall Trust Grade A (TM ≥ 100) and live verified blob evidence (The Star Bar, META.md §2.4).
    • Unique Branch (standalone, no suiteComponents): 4★ Unique requires bucket-level Origin on its generic bucket (META.md §4.1).
    • Suite Branch (capstone carrying suiteComponents): 4★ Extra does NOT require bucket-level Origin on its generic bucket. The capstone represents the suite cluster rather than a single generic bucket monopoly. Suite origin requirements apply at 5★ Ultimate (proposer must hold Origin on ≥1 of the suiteComponents per META.md §4.2).
    • Precedents: leonxlnx/taste-skill (4★ Extra, origin: false on generic design-generation), gsd-build/get-shit-done (4★ Extra).

Inputs

Gather these before beginning:

InputDescriptionExample
contributorGitHub organization or usernameleonxlnx
suite-slugIdentifier for the suite capstonetaste-skill
repo-urlCanonical GitHub repository URLhttps://github.com/leonxlnx/taste-skill
default-branchDefault git branch of the upstream repomain
componentsList of sub-skill directories / filesbrandkit, minimalist-skill, ...

Phase 1: Upstream Suite Inspection & Installability Audit

Inspect the upstream repository structure and audit every component before touching the registry.

1.1 Repo Structure & Component Identification

Inspect the remote repository using gh or clone locally in a temporary directory:

gh repo view <owner>/<repo> --json defaultBranchRef,stargazerCount,description

Locate all component SKILL.md files:

  • Typical layouts: skills/<component-name>/SKILL.md, plugins/<name>/SKILL.md, or top-level directories.
  • Identify the capstone: the primary entry point, meta-orchestrator, or root skill that references or coordinates the components.

1.2 Frontmatter & Lifecycle Audit

For each component SKILL.md, verify:

  • Frontmatter exists with valid YAML fences (---).
  • Required fields: name, description.
  • Instructions / CLI usage: verify how the skill is executed (e.g. npx, uvx, shell script, or harness prompt).

1.3 The Star Bar (Installability Gate)

Per META.md §2.4:

  • Any skill promoted to 3★ or higher MUST have a verified GitHub BLOB link pointing to a concrete file (e.g. https://github.com/<owner>/<repo>/blob/<branch>/.../SKILL.md), never a bare repo root.
  • A missing verified blob link is a hard reset to 1★ (Awakened).
  • Construct the exact blob URL for every component and confirm HTTP 200 reachable.

Phase 2: Origin Candidate Analysis & Generic Mapping

Map every component (and the capstone) to a generic capability bucket in registry/nodes/.

2.1 Capability Resolution

For each component:

  1. Search existing generic nodes:
    grep -rn "name: " registry/nodes/
    # or search by id slug
    find registry/nodes -name "*.json" | grep -i "<keyword>"
    
  2. If a matching generic bucket exists:
    • Check existing named implementations in that bucket:
      grep -rl "genericSkillRef: <generic-id>$" registry/named/
      
    • If bucket has no implementations, it is vacant. The incoming component is a primary Origin candidate.
    • If bucket has an incumbent: evaluate whether the incoming component is demonstrably more renowned. If not, incoming component will have origin: false.
  3. If no matching generic capability exists:
    • Create a starless generic basic node:
      gaia dev add <new-generic-id> \
        --name "<Generic Capability Name>" \
        --description "<Precise, vendor-agnostic capability description>" \
        --type basic \
        --no-build
      
    • The incoming component is the sole implementation and therefore an Origin candidate.

Phase 3: Ingestion of Named Components & Evidence Scoring

Ingest components programmatically via gaia dev verbs. Never hand-edit JSON or frontmatter.

3.1 Register Named Skills

For each component:

gaia dev add --named <contributor>/<component-slug> \
  --generic-ref <generic-id> \
  --name "<Component Display Name>" \
  --title "<RPG Title / Clean Name>" \
  --description "<Component description>" \
  --links github="https://github.com/<owner>/<repo>/blob/<branch>/<path>/SKILL.md" \
  --no-build

Also register the named capstone:

gaia dev add --named <contributor>/<suite-slug> \
  --generic-ref <capstone-generic-id> \
  --name "<Suite Capstone Display Name>" \
  --title "<Suite Title>" \
  --description "<Suite overview description>" \
  --links github="https://github.com/<owner>/<repo>/blob/<branch>/<path-or-readme>" \
  --no-build

3.2 Add Verified Evidence Rows

Gather verified repository metrics from GitHub API:

  • stargazerCount: Total stars on the suite repository.
  • commits: Total commit count.
  • contributors: Total contributor count.
  • createdAt / sourceStartedAt: Creation timestamp of repo/file.

A. Capstone Evidence

Add repository adoption evidence to the capstone:

gaia dev evidence <contributor>/<suite-slug> \
  --type github-stars-own \
  --source "https://github.com/<owner>/<repo>/blob/<branch>/<path>" \
  --stars <stargazerCount> \
  --source-started-at "<YYYY-MM-DD>" \
  --notes "<stars> stars on <owner>/<repo> repository" \
  --no-build

gaia dev evidence <contributor>/<suite-slug> \
  --type repo-own \
  --source "https://github.com/<owner>/<repo>" \
  --commits <commitCount> \
  --contributors <contributorCount> \
  --source-started-at "<YYYY-MM-DD>" \
  --notes "<commits> commits, <contributors> contributors" \
  --no-build

B. Component Evidence & Shared Repo Cap

Add component-level evidence. Sub-components residing in the shared repository inherit adoption signals, but Yggdrasil III enforces:

  • skillCountInRepo: Set to the total number of components sharing the repository.
  • SUITE_COMPONENT_REPOSITORY_CAP = 50.0 TM: The scoring engine automatically bounds shared repository evidence so that adoption alone caps at 50 TM (Grade B).
gaia dev evidence <contributor>/<component-slug> \
  --type github-stars-own \
  --source "https://github.com/<owner>/<repo>/blob/<branch>/<path>/SKILL.md" \
  --stars <stargazerCount> \
  --skill-count <componentCount> \
  --source-started-at "<YYYY-MM-DD>" \
  --no-build

Phase 4: Suite Manifestation via gaia dev fuse

Critical Rule: Running gaia dev update-named --suite-components alone is insufficient. gaia dev docs and gaia dev build will strip suiteComponents from Named Skill frontmatter unless a backing suite manifest exists at registry/suites/<contributor>/<suite-slug>.json.

Use gaia dev fuse to establish the manifest and wire the graph in one atomic operation:

gaia dev fuse <generic-suite-id> \
  --name "<Suite Name>" \
  --description "<Suite fusion description>" \
  --prereqs <comp1-generic-id>,<comp2-generic-id>,... \
  --named-capstone <contributor>/<suite-slug> \
  --suite-components <contributor>/<comp1>,<contributor>/<comp2>,... \
  --no-build

Verification:

  1. Confirm manifest exists: test -f registry/suites/<contributor>/<suite-slug>.json
  2. Confirm frontmatter wiring:
    grep -c "suiteComponents:" registry/named/<contributor>/<suite-slug>.md
    

Phase 5: Calibration of Capstone & Component Origins

5.1 Appraise Trust Magnitude

Run the appraisal tool on the capstone and all components:

python scripts/trust_appraise.py --skill <contributor>/<suite-slug>

Or use the detailed inspector:

GAIA_OPERATOR_OVERRIDE=1 PYTHONPATH=src python3 scripts/inspectTrustMagnitude.py --skill <contributor>/<suite-slug>

Verify:

  • Capstone TM score and Overall Trust Grade.
  • fusionScore: Structural composition reading (informational, does not feed TM).

5.2 Capstone Calibration (4★ Extra)

Evaluate the promotion gate:

  • Condition: Capstone has Overall Trust Grade A (TM ≥ 100) and verified blob link.
  • Suite Branch Invariant: Capstone does NOT require bucket-level Origin on its generic bucket! Origin on generic bucket is strictly for Unique branch standalone skills (META.md §4.1). Suite origin requirements apply at 5★ Ultimate (Origin on ≥1 suiteComponent per META.md §4.2).
  • Apply calibration:
    gaia dev calibrate <contributor>/<suite-slug> 4★ --no-build
    

5.3 Component Calibration & Origin Assignment

  1. Origin Assignment: For components that are the sole or most renowned implementation of their generic bucket:
    gaia dev update-named <contributor>/<component-slug> --origin true
    
  2. Component Star Calibration: Calibrate each component to its TM grade band:
    • Grade B (TM ≥ 50) → 3★ (must have verified blob link)
    • Grade C (TM ≥ 20) → 2★
    • Ungraded (TM < 20) → 1★
    gaia dev calibrate <contributor>/<component-slug> <N★> --no-build
    

Phase 6: Single-PR Closeout with Class S Documentation Generation

6.1 Single Dev Build

Run exactly one build pass to compile the registry and generate Class S site assets:

GAIA_OPERATOR_OVERRIDE=1 PYTHONPATH=src python3 -m gaia_cli.main dev build

6.2 Preflight Hygiene & Staging

Use scripts/review_meta_close.py to prevent leaks, drop CRLF noise, and stage only allowlisted artifacts:

# 1. Review status (must show 0 leaks in PR)
python scripts/review_meta_close.py status -v

# 2. Stage intended artifacts for this contributor
python scripts/review_meta_close.py stage --contributor <contributor> --apply

# 3. Verify clean index
python scripts/review_meta_close.py check

# 4. Run UTF-8 safe validation
python scripts/review_meta_close.py validate

6.3 Staged Artifacts Checklist

Verify that git diff includes only:

  • registry/nodes/ (new/updated generic nodes)
  • registry/named/<contributor>/ (capstone and component .md files)
  • registry/suites/<contributor>/<suite-slug>.json (suite manifest)
  • registry/named-skills.json & tracked registry files
  • docs/graph/ (Class S graph assets: gaia.json, gaia.gexf, named/index.json)
  • docs/badges/_assets/<contributor>/ & docs/og/<contributor>/

6.4 Branch & PR Opening

git checkout -b review/meta/<contributor>-<suite-slug>
git commit -m "feat(suite): ingest <contributor>/<suite-slug> suite with <N> components"
git push origin review/meta/<contributor>-<suite-slug>
gh pr create --title "feat(suite): ingest <contributor>/<suite-slug> suite" \
  --body "Ingests <contributor>/<suite-slug> suite under Yggdrasil III. Calibrates capstone to 4★ Extra on Grade A TM."

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

ci-churn

無料

Measure CI churn cost for a pull request: how many commits and how many CI-compute seconds were burned on avoidable "fix the CI" push rounds. Use when you want to understand the efficiency cost of a PR's iteration cycle, get actionable pre-push check suggestions, or include a churn summary in a pipeline report. Triggers: "ci churn", "how much CI did we burn", "avoidable commits", "ci efficiency", "wasted CI time", "churn report", "/ci-churn". Called automatically by /fp-drift at pipeline close.

日本語の概要は準備中です。原文の説明を表示しています。

gaia-research/gaia-skill-tree232026年10月13日 更新

Produce the rendered evidence a human-gated frontend PR owes its reviewer: capture each changed surface (plus the before state for anything modified, plus mobile width), then generate one self-contained local HTML page the founder can open and review without checking anything out. Use when a PR is gated by the "Frontend changes are HUMAN-GATED" section of CLAUDE.md, or when the user says "/design-gate", "screenshot evidence for this PR", "prep this for founder review", "build the design evidence page", "what does this look like before and after". The page comes from a template script, not hand-written markup. Output is session-local and never committed. Do NOT use for non-gated changes: typo/grammar/factual fixes in an unchanged block, dead link targets, alt/ARIA fixes with no visual change, cache-bust bumps, or refactors with no rendered diff.

日本語の概要は準備中です。原文の説明を表示しています。

gaia-research/gaia-skill-tree232026年10月13日 更新

One-fix-at-a-time, approval-gated design fix recovery. Given a checklist of fixes (handover doc, CHECKLIST.md, or inline list), applies each individually, presents a diff, and waits for an explicit APPROVE / REJECT / NO CHANGES decision before committing. Trigger phrases: /design-iteration, "go through the checklist one by one", "fix these one at a time", "bucket recovery", "design fix pass".

日本語の概要は準備中です。原文の説明を表示しています。

gaia-research/gaia-skill-tree232026年10月13日 更新

Playbook for correcting a named skill's star level to match its computed Trust Magnitude grade, with an independent post-calibration verification pass and a CLI-logged timeline event (never hand-written). Use when: recalibrating a skill's level, fixing a stale or drifted star rating, running a batch of `gaia dev calibrate` corrections, or confirming a past calibration actually landed. Triggers: "calibrate <skill>", "recalibrate the trust magnitude backlog", "promote/demote <skill> to N stars", "why is <skill> still N stars", "verify a calibration actually landed", "Hall of Heroes ranking is wrong".

日本語の概要は準備中です。原文の説明を表示しています。

gaia-research/gaia-skill-tree232026年10月13日 更新

Close out a completed EPIC / sprint end-to-end: author a fact-verified meta-post announcing what shipped, archive the EPIC's handovers, snapshot orchestrator memory, gate the integration→main merge on CI-green + explicit human go, drive the version bump (merge-commit conventions that auto-sync classifies), and verify the release reached the registry/PyPI. Use when the user says "close out EPIC <n>", "wrap up the sprint", "ship the meta shift", "/epic-close", or asks to finalize an integration branch that is ready to merge to main. Generalizes the EPIC 1002 (Yggdrasil II) close-out. This skill orchestrates — it authors founder docs directly (superadmin) and delegates code/regen to workers; it NEVER merges to main without the human gate.

日本語の概要は準備中です。原文の説明を表示しています。

gaia-research/gaia-skill-tree232026年10月13日 更新

Run this skill for Phase 3 of the evidence verification pipeline — the adversarial audit step. Use when you need to check the evidence data lake for bad data before ingestion: dead links, wrong URL formats (tree/ vs blob/), subjective wording ("elite", "high-quality"), stale migration notes, benchmark catalog misuse/vendor-claim leakage, or skills whose star evidence conflicts with classified evidence level. Triggers on phrases like: "audit the data lake", "adversarial check", "ev-adversarial-audit", "check for noise in evidence", "flag bad evidence", "run the audit phase", "quality check the by-type files", or any reference to Phase 3 of the pipeline.

日本語の概要は準備中です。原文の説明を表示しています。

gaia-research/gaia-skill-tree232026年10月13日 更新

gaia-research のスキルをすべて見る

このスキルの問題を報告する