By potenlab
Dual knowledge-graph workflow built on understand-anything: Impl KG (autoUpdate, derived from source) + SSOT KG (intent/spec, projected from a Decision Log). Bootstraps via /kg-init.
Pre-flight context dispatcher invoked by the main Claude session before any substantive code or architecture work. Classifies the prompt, fans out to kg-ssot-check / kg-impl-check / kg-history-check in PARALLEL, synthesizes their findings, and returns a short briefing plus a SIDE-EFFECT WARNING when the user's proposed work would breach a node's touch_budget or affect dependents that haven't been acknowledged. Invoke this every time you're about to make a non-trivial code change or design decision.
Search prior user prompts, decisions, and sessions for context relevant to the current prompt. Uses the `entire` CLI (entire-search) to query the local checkpoint store, and the understand-anything skills to relate findings back to KG nodes. Invoked in parallel by kg-context-dispatch — do not call directly.
Look up what the Impl KG (`.understand-anything/knowledge-graph.json`) says about the area the user is touching. Returns matching nodes + their incoming dependents so the dispatcher can compute side-effect risk. Invoked in parallel by kg-context-dispatch — do not call directly.
Look up what the SSOT KG (`.understand-anything-ssot/knowledge-graph.json`) says about the area the user is touching. Returns SSOT node fields (status, acceptance, contract, rationale_ref, touch_budget) for every node relevant to the user's prompt. Invoked in parallel by kg-context-dispatch — do not call directly.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A Claude Code plugin that bootstraps a dual knowledge-graph + parallel-context-dispatch workflow in any repo. Built on top of understand-anything and Entire.
Run /kg-init once. You get:
.understand-anything/knowledge-graph.json — what the code IS. Maintained by understand-anything..understand-anything-ssot/knowledge-graph.json — what the code SHOULD BE. Seeded once from the Impl KG with SSOT defaults (status, acceptance, contract, rationale_ref, touch_budget)..entire/ — local prompt/decision history, automatically enabled if the entire CLI is present.kg-context-dispatch, kg-ssot-check, kg-impl-check, kg-history-check — wired into CLAUDE.md so every substantive prompt gets a pre-flight context briefing and a side-effect warning.scripts/ssot_seed.py for deterministic re-seeding if you ever need it.Every substantive prompt (not greetings or "looks good") triggers this flow:
User prompt
│
▼
Main Claude reads CLAUDE.md → Task(kg-context-dispatch)
│
▼
kg-context-dispatch (Phase A: cheap classify; Phase B: extract topic hints)
│
▼ Phase C: fan out IN PARALLEL — single message, three Task calls
├──────────────────────────┬──────────────────────────┐
▼ ▼ ▼
kg-ssot-check kg-impl-check kg-history-check
.understand-anything- .understand-anything/ .entire/ via
ssot/ → status, → relevant nodes + entire search →
acceptance, contract, DEPENDENTS (the prior decisions,
touch_budget, side-effect signal) sessions, prompts
rationale_ref
│ │ │
└──────────────────────────┴──────────────────────────┘
│
▼
kg-context-dispatch (Phase D: synthesize + compute side_effect_warning)
│
▼
Main Claude proceeds with compact JSON briefing as authoritative context.
If side_effect_warning=true, surface warning_text to user BEFORE editing.
kg-context-dispatch flags side_effect_warning: true when:
touch_budget (from SSOT).deprecated or drift.The warning cites specific node IDs / file paths / DL IDs — not hand-wavy "this might affect things".
Sub-agents have isolated context windows. Main Claude pays for:
The three leaf agents (read full KG fragments, run entire search, etc.) don't bill against the main session's context. And because they run in one parallel message, wall-clock = max(3) not sum(3).
The dispatcher's Phase A classifier skips trivial prompts entirely — you don't pay the 4-agent cost on every "ok" or "show me the README".
kg-workflow bootstraps and wires the dispatch flow. It does not:
docs/ssot/ or any process directory (kg-init prompts the user if no such dir is detected)./plugin marketplace add potenlab/kg-workflow
/plugin install kg-workflow
After install, run /clear so the four agents register in the session.
curl -fsSL https://raw.githubusercontent.com/potenlab/kg-workflow/main/install.sh | bash -s <platform>
Supported platforms: codex, gemini, opencode, vscode, vibe. See install.sh --help.
understand-anything plugin installed (kg-workflow calls its /understand skill to build the Impl KG, and its agents are useful for KG lookup).python3 3.11+ on the user's machine.kg-init refuses to run outside one.docs/ssot/). If missing, kg-init stops and prompts — see below.entire CLI from https://docs.entire.io. Without it, kg-history-check returns entire_not_initialized on every dispatch (graceful degrade, not fatal).cd your-repo
/kg-init
kg-init will, in order:
npx claudepluginhub potenlab/kg-workflow --plugin kg-workflowFull-stack project planning and execution plugin for Claude Code. Orchestrates specialist agents to plan, build, and iterate on projects.
Agentic SaaS validation flow: brand intake → Higgsfield logo + moodboard → approval gate → animated landing page with waitlist. Powered by the Higgsfield MCP server.
Describe an app in plain language, get a running app + PR. A designer or PM speaks product; Claude does the full-stack engineering underneath (data model -> backend -> UI -> running app -> PR) and only ever shows a working app, never code. Auto-triggers on app-creation intent; builds inside an isolated sandbox that never touches production data, secrets, or auth/billing.
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Harness-native ECC plugin for engineering teams - 67 agents, 279 skills, 94 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, plus codex, antigravity, and grok CLIs when installed) to get diverse perspectives on coding problems
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.