Persistent project state + unified skill registry for stateless LLMs (Grok, OpenClaw, and compatible agents). Eliminates amnesia between turns.
How this skill is triggered — by the user, by Claude, or both
Slash command
/grok-persistent-state:grok-persistent-stateThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Primary use:** When working with Grok (or any stateless LLM) inside OpenClaw, Claude Code, Cursor, or custom agent loops and you need durable memory + a single unified capability surface across sessions.
Primary use: When working with Grok (or any stateless LLM) inside OpenClaw, Claude Code, Cursor, or custom agent loops and you need durable memory + a single unified capability surface across sessions.
Use this skill automatically or when the user says any of:
Grok (and most LLM APIs) are stateless. Every request is independent. Without external memory:
This skill restores persistent project state and an omniskill-style unified registry that survives across requests.
Durable State (machine + human readable)
project_state.json — canonical structured memoryPROJECT_STATE.md (optional) — human-readable summaryUnified Tool/Skill Surface
skills/<name>/manifest.jsonSession Lifecycle
patch_state with deltas{
"project_id": "default",
"goal": "...",
"current_focus": "...",
"recent_changes": ["..."],
"open_tasks": ["..."],
"decisions": ["..."],
"artifacts": ["path/to/file"],
"skill_preferences": {},
"updated_at": "ISO8601"
}
At the start of any user request:
read_project_state (or load from disk).After any meaningful change (code written, decision made, task completed):
patch_state(project_id, {
"current_focus": "new focus",
"recent_changes": [...previous, "what just happened"],
"open_tasks": [remaining tasks],
"decisions": [...],
"artifacts": [new files]
})
Never let the LLM hallucinate prior context when state is available.
After successful completion of a step:
This skill MUST:
write_project_state / patch_stateThis skill MUST NEVER:
Data that stays local: Everything. State lives in the user's project directory.
Data that leaves the machine: Only what the underlying LLM (Grok, etc.) would normally see in the prompt.
User: "Continue implementing the revenue dashboard. We were working on the Stripe sync last time."
Agent (with this skill):
current_focus was "Stripe billing sync", open tasks include "add retry + idempotency".Extracted and generalized from Igor Holt's clawbot-grok reference implementation (persistent state + skill_loader patterns) and Genesis Conductor thermodynamic agent orchestration work.
Use this skill whenever long-running, stateful agent work with Grok or OpenClaw is required.
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
npx claudepluginhub igor-holt/openclaw-skills --plugin grok-persistent-state