Name: livebase
Author: Yuzc-001

Imported from Yuzc-001/livebase (SKILL.md). Install upstream with npx skills add Yuzc-001/livebase. Copyright stays with the author.

Livebase

Use Livebase when the problem is not "remember more" but "continue work correctly."

The center is:

resume real work from the last unresolved evidence gap

Core Rules

Resume before acting.
Checkpoint when work pauses, blocks, or hands off.
Generate a contract before bounded execution.
Write back a verification ledger, not an essay.
Persist only the smallest durable residue.

Command Loop

bun run setup
bun run task create "Title" --goal "..."
bun run checkpoint <task_id> --evidence-gap "question::suggested action"
bun run resume <task_id>
bun run contract <task_id>
bun run writeback

If the host supports MCP, prefer the tool surface:

livebase.task.resume
livebase.task.contract
livebase.task.writeback
livebase.task.checkpoint
livebase.task.resolve_gap

Host-side prompt guidance lives in docs/host-adapter-prompt-contract.md. Copy-paste host presets live in docs/adapter-presets/README.md.

When To Use

A recurring task should continue tomorrow without re-reading everything.
An agent is about to start work and needs a bounded contract.
A task is blocked and the exact evidence gap needs to be preserved.
Completed work should improve the next run through residue.

When Not To Use

You only need a generic note-taking tool.
You want to dump full transcripts into storage.
You are looking for a broad chat memory system.

Agent Behavior

Prefer source-linked grounding over free-floating summaries.
Preserve the exact evidence gap that blocks progress.
Use resume to recover the next step, not to fetch broad history.
Use contract to define boundaries before external execution.
Use writeback only with observed checks, remaining ambiguity, and durable residue.
Do not save every thought. Save what will make the next run better.

Install livebase and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "yuzc-001-livebase-livebase", "kind": "skill", "name": "livebase", "description": "Local-first work continuity engine for humans and agents. Use when work should continue from the last unresolved evidence gap instead of restarting from zero, especially when you need to checkpoint a task, resume it later, generate an execution contract, or write back a verification ledger.", "publisher": "Yuzc-001", "version": "1.0.0", "capabilities": { "domains": [ "legal", "hr" ], "tags": [ "skill-md", "github" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "Local-first work continuity engine for humans and agents. Use when work should continue from the last unresolved evidence gap instead of restarting from zero, especially when you need to checkpoint a task, resume it later, generate an execution contract, or write back a verification ledger." ], "primary": false, "metadata": { "source": { "provider": "github", "repository": "https://github.com/Yuzc-001/livebase", "path": "SKILL.md", "ref": "72cf6239e3dd5b329defb779fc68faca049e70fa", "url": "https://github.com/Yuzc-001/livebase/blob/72cf6239e3dd5b329defb779fc68faca049e70fa/SKILL.md", "key": "Yuzc-001/livebase/SKILL.md" } }, "instructions": "# Livebase\n\nUse `Livebase` when the problem is not \"remember more\" but \"continue work correctly.\"\n\nThe center is:\n\n> resume real work from the last unresolved evidence gap\n\n## Core Rules\n\n1. Resume before acting.\n2. Checkpoint when work pauses, blocks, or hands off.\n3. Generate a contract before bounded execution.\n4. Write back a verification ledger, not an essay.\n5. Persist only the smallest durable residue.\n\n## Command Loop\n\n```bash\nbun run setup\nbun run task create \"Title\" --goal \"...\"\nbun run checkpoint <task_id> --evidence-gap \"question::suggested action\"\nbun run resume <task_id>\nbun run ", "cost": { "context_tokens": 482 } }

livebase

About

Livebase

Core Rules

Command Loop

When To Use

When Not To Use

Agent Behavior

Install livebase and it becomes one more candidate the router can pick - when it fits.

Use it

Manifest

Reviews