graft-work-intake

name	graft-work-intake
description	Repository-specific workflow orchestrator for new long-running work. Use when startup discovers a request that may need design, topic, roadmap, ADR, bootstrap, or loop dispatch, and the repository must classify it through one Work Intake path instead of special-case entry flows.

Graft Work Intake

Use this skill as the only workflow-level entry for new long-running work in Graft.

Treat root AGENTS.md, ai-plan/AGENTS.md, ai-plan/README.md, and the approved ai-plan ADRs as the authority sources. This skill is a thin orchestration wrapper: it does not define independent business rules, startup rules, or content-generation rules.

Guardrail summary:

• do not define independent business rules

Read First

1. Complete the root AGENTS.md startup preflight.
2. Read ai-plan/AGENTS.md.
3. Read ai-plan/README.md.
4. Read ai-plan/design/governance/ai/AI任务追踪与恢复设计.md.
5. Read ai-plan/design/governance/ai/AI工具与MCP接入治理规范.md.
6. Read ai-plan/design/decisions/ADR-003-work-intake-and-bootstrap-model.md.
7. Read ai-plan/public/README.md only after startup preflight and only to confirm whether an active topic already owns the work.

When To Use

Use this skill when all of the following are true:

• the request introduces new work rather than resuming an already-owned active topic
• the work may be long-running, multi-batch, or recovery-sensitive
• the repository needs to decide whether it requires design, topic, roadmap, ADR, bootstrap, or loop dispatch

Typical triggers:

• design a new capability
• add a multi-batch feature
• plan a refactor that may need topic tracking
• create the long-running work entry for this request

Workflow

1. Confirm whether the request is short-lived or long-running.
2. If the work is short-lived, do not create a new topic or persisted Work Contract; hand off directly to normal implementation flow.
3. If the work is long-running, produce one Work Contract.
4. Decide the contract fields from documented rules only:
   • kind
   • scope
   • authority_summary
   • requires.design
   • requires.topic
   • requires.roadmap
   • requires.adr
   • execution.engine
   • execution.dispatch_skill
   • bootstrap.targets
   • closeout.archive
   • closeout.lessons_review
5. Apply contract-driven minimal bootstrap:
   • create only the minimal skeleton required by the contract
   • do not pre-create artifacts the contract did not request
   • do not fill in domain content beyond starter structure
6. Persist the Work Contract only when the contract creates an active topic:
   • store it in the topic tracking file
   • do not create a standalone work-contract.yaml
   • do not put the full contract in topic README.md
   • do not put the full contract in ai-plan/catalog.json
7. Dispatch the work to specialized skills after bootstrap:
   • design content goes to the relevant design or governance skill
   • long-running execution normally goes to graft-multi-agent-loop
   • closeout still goes through graft-task-closeout

Guardrails

• do not create a second startup path, second validation truth, or second recovery store
• do not treat design, roadmap, topic, or ADR as peer entry workflows
• do not author domain content inside this skill
• do not persist a Work Contract for short-lived work
• do not let bootstrap guess future artifacts beyond what the contract explicitly requires

Closeout

Report:

work_intake:
- work_classification:
- work_contract_created:
- persisted_contract:
- bootstrap_targets:
- dispatch_skill:
- existing_topic_reused: