{
  "markdown": "# Ark Skills Plugin\n\nClaude Code plugin providing 6 shared skills to all ArkNode projects, built around two co-equal cores: a **workflow consultant** that plans and routes each task to one execution ecosystem, and a **conventions layer** that converges projects onto shared OKF knowledge bundles + GitHub-Issues task management. A context-discovery pattern adapts every skill to the active project at runtime.\n\n## Installation\n\n```bash\n# Add the marketplace\n/plugin marketplace add HelloWorldSungin/ark-skills\n\n# Install the plugin (user-scoped — available in all projects)\n/plugin install ark-skills@ark-skills\n\n# Verify skills are available\n/ark-health\n```\n\n### Prerequisites\n\n**Required for all skills:** None — skills are instruction-only.\n\n**Optional — enhances task tracking, retrieval, and routing:**\n\n| Dependency | Skills Enhanced | Install |\n|------------|----------------|---------|\n| [`gh` CLI](https://cli.github.com) | `/ark-consult`, `/ark-onboard` (GitHub-Issues task management) | `brew install gh` + `gh auth login` |\n| [mattpocock/skills](https://github.com/mattpocock/skills) | `/ark-consult` (issue machinery + a routing destination) | `npx skills@latest add mattpocock/skills` + `/setup-matt-pocock-skills` |\n| [NotebookLM CLI](https://github.com/nichochar/notebooklm-cli) | `/notebooklm-vault` (synthesized recall) | `pipx install notebooklm-cli` + `notebooklm login` |\n\nRouting destinations for `/ark-consult` (install per your workflow): superpowers, gstack, oh-my-claudecode.\n\n## Available Skills\n\n| Skill | Core | Description |\n|-------|------|-------------|\n| `/ark-consult` | Workflow consultant | Stateless planning phase for any non-trivial task: triages, recommends exactly ONE execution ecosystem from a routing matrix, files a GitHub epic as the plan of record, and hands off. |\n| `/vault` | Conventions | Write-side OKF knowledge ops for the bundle at `vault/`: end-of-session distillation, document ingestion, `log.md` append, index regen. |\n| `/notebooklm-vault` | Conventions | Synthesized-recall backend: syncs the OKF bundle to NotebookLM and answers factual/historical queries. |\n| `/ark-onboard` | Conventions | Interactive setup wizard: OKF bundle init, label bootstrap, mattpocock setup, CLAUDE.md rows. |\n| `/ark-health` | Conventions | Diagnostic scorecard for the v2 invariants. |\n| `/ark-update` | Conventions | Version-driven migration framework; converges downstream projects onto the current target profile (execution deferred per-project). |\n\n## Skill Documentation\n\n**`/ark-consult`** — Given a non-trivial task, triages it, asks at most two clarifying questions, and recommends exactly ONE execution ecosystem (gstack / superpowers / mattpocock / oh-my-claudecode) from a 12-archetype routing matrix with reasoning. Files a GitHub epic (epic + component + priority labels; children as task-list checkboxes, optionally cut via mattpocock `/to-tickets`), then invokes the chosen ecosystem's entry skill and stops. Never chains two ecosystems, never orchestrates post-handoff, never writes local state — the GitHub epic is the only state (resume = `gh issue view <epic>`).\n\n**`/vault`** — Write-side knowledge operations for the OKF bundle: distills durable insights from a session (knowledge only — no session logs, no task tracking), ingests external documents, appends work-record lines to `log.md`, and always finishes by regenerating the index and passing `okf_lint.py`. Reading the vault is script-based (`okf_cli.py`) or via NotebookLM.\n\n**`/notebooklm-vault`** — Bridges the OKF bundle with Google NotebookLM for synthesized cross-session recall. Sub-commands: `setup`, `ask`, `session-continue` (resumes from the `log.md` tail + the referenced GitHub epic), `bootstrap`, `audio`, `report`, `conflict-check`, `status`.\n\n**`/ark-onboard`** — Interactive setup wizard converging a project onto the v2 conventions: OKF bundle init, GitHub-Issues label taxonomy + `docs/agents/` conventions, mattpocock setup, and CLAUDE.md rows. Handles greenfield, migration, repair, and healthy states. Exempt from context-discovery.\n\n**`/ark-health`** — Runs the v2 invariant checks across five areas (routing-destination ecosystems, the 6-skill roster, OKF conventions, GitHub-Issues conventions, NotebookLM recall) and produces a scored scorecard with fixes. No auto-fix — points to `/ark-onboard` for remediation. Exempt from context-discovery.\n\n**`/ark-update`** — Version-driven migration framework. Converges downstream projects onto the current target profile; its v2.0.0 profile declares two pending migrations (OKF conversion + GitHub-Issues adoption), executed per-project later. Distinct from `/ark-onboard` repair (failure-driven).\n\n## Context-Discovery Pattern\n\nEvery skill uses **context-discovery** — no skill contains hardcoded project names or vault paths. When invoked, each skill reads the project's `CLAUDE.md`, follows the monorepo hub link if present, and extracts the project name, vault root, and project docs path. See `CLAUDE.md` in this repo for the full procedure.\n\n**Exemption:** `/ark-onboard` and `/ark-health` are exempt — they must work when CLAUDE.md is missing, broken, or incomplete.\n\n## Architecture\n\n```\nark-skills (Claude Code plugin)\n├── .claude-plugin/\n│   ├── plugin.json           # Plugin metadata\n│   └── marketplace.json      # Repo-level plugin registry\n├── skills/                   # 6 shared skills\n│   ↓ context-discovery\n│   Project CLAUDE.md → vault root, project docs path\n├── docs/agents/              # GitHub-Issues conventions (issue-tracker, triage-labels, domain)\n└── vault/                    # OKF v0.1 knowledge bundle (_meta/okf tooling, index.md, log.md)\n        ↓\n   NotebookLM sync (synthesized recall)\n   GitHub Issues (work record)\n```\n\n## New Project Onboarding\n\nRun `/ark-onboard` for interactive guided setup. It detects your project state and walks you through the appropriate path (greenfield / migration / repair / healthy).\n\n## Repository Structure\n\n| Directory | Purpose |\n|-----------|---------|\n| `.claude-plugin/` | Plugin manifest (plugin.json, marketplace.json) |\n| `skills/` | 6 shared skill definitions (SKILL.md files) |\n| `docs/agents/` | GitHub-Issues conventions (issue-tracker, triage-labels, domain) |\n| `docs/` | Design specs, plans, historical archives |\n| `vault/` | The plugin's own OKF v0.1 knowledge bundle |\n\n## Development\n\n### Modifying Skills\n\n1. Edit `skills/<skill-name>/SKILL.md`.\n2. Test by invoking the skill from a project.\n3. Verify no hardcoded references: `rg -n \"ArkPoly|ArkSignal|trading-signal-ai|CT100|192\\.168\" skills/`.\n4. Commit and push (bump `VERSION`, `plugin.json`, `marketplace.json`, `CHANGELOG.md` in the same commit).\n\n### Verification Checks\n\n```bash\n# Exactly 6 skills\nfind skills -mindepth 1 -maxdepth 1 -type d ! -name shared | wc -l  # → 6\n\n# OKF bundle lints clean\npython3 vault/_meta/okf/okf_lint.py --quiet; echo $?  # → 0\n\n# Every skill references context-discovery (except the two exempt onboarding skills)\nrg -L \"Project Discovery|CLAUDE.md|Context-Discovery\" skills/*/SKILL.md\n```\n",
  "bytes": 7059,
  "sha": "f3ddc5edcc3c95e934833b7f53ff7a03953fa779b95542b6d5816635e30a1de3",
  "repo_slug": "helloworldsungin/ark-skills",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/okf_helloworldsungin_ark_skills_vault_index__dba1b82f/readme"
}