| name | ll-gate-human-orchestrator |
| description | Governed LL workflow skill for orchestrating human gate review over a gate_ready_package, producing authoritative brief, pending-human, decision, dispatch, and approve-auto-materialization outcomes through the mainline gate runtime. |
LL Gate Human Orchestrator
This skill is the governed gate workflow for turning one gate_ready_package into one authoritative gate decision package. It wraps the real ll gate evaluate, ll gate dispatch, and ll gate materialize runtime surface instead of inventing a parallel gate engine.
Canonical Authority
- Governing ADRs:
E:\ai\LEE-Lite-skill-first\ssot\adr\ADR-016-Gate Skill 作为第二会话 Human Gate Orchestrator 的实现基线.MD, ADR-015, ADR-001
- Upstream runtime object:
gate_ready_package
- Primary runtime command:
python skills/ll-gate-human-orchestrator/scripts/gate_human_orchestrator.py run --input <gate-ready-dir> --repo-root <repo-root>
- Real execution surface:
ll gate evaluate, ll gate dispatch, ll gate materialize
Runtime Boundary Baseline
- Classify this gate workflow under
E:\ai\LEE-Lite-skill-first\ssot\adr\ADR-038-运行时核心抽象边界与对象分层基线.MD.
- This skill is the governed
Skill and Workflow wrapper for human gate orchestration. The gate runtime commands are carriers; Gate remains the state-advance authority.
Required Read Order
ll.contract.yaml
input/contract.yaml
output/contract.yaml
agents/executor.md
agents/supervisor.md
input/semantic-checklist.md
output/semantic-checklist.md
Execution Protocol
- Accept only a governed
gate_ready_package directory or the canonical gate-ready-package.json file.
- Validate the package structure before issuing any gate decision.
- Use the mainline gate runtime to build
gate-brief-record, gate-pending-human-decision, and gate-decision.
- Treat
approve as approval plus immediate materialization on the default runtime path.
- Always dispatch the authoritative decision after evaluation.
- Record execution evidence and then require a supervisor pass before freeze.
- Freeze only after
validate-package-readiness returns success.
Human Interaction Rule
- When using
claim-next, prepare-round, or show-pending, do not stop at file paths or raw JSON.
- Always surface the approval context directly in the reply:
- what object is under review
- what decision is being requested
- the main focus points
- the allowed reply formats
- Default behavior: if the runtime returns
human_brief, render the full markdown inline in the first reply for that pending item.
- Do not ask whether to paste the brief first; paste it by default unless the user explicitly asked for a shorter summary-only mode.
- Do not collapse to only
review_summary or a short abstract unless the user explicitly asks for a shorter version.
- If the runtime returns
review_summary, summarize the key fields in plain Chinese before asking for a decision.
- Default the human-facing brief to Chinese unless the user explicitly asks for another language.
- The user should not need to open artifact files just to understand the pending approval item.
Workflow Boundary
- Input: one
gate_ready_package
- Output: one
gate_decision_package
- Out of scope: upstream candidate authoring, FEAT/TECH derivation, and replacing the underlying
ll gate ... runtime
Non-Negotiable Rules
- Do not bypass
ll gate evaluate with handwritten decision files.
- Do not treat human decision actions as runtime states.
- Do not require a second skill for materialization after
approve.
- Do not accept raw requirements or standalone markdown as gate input.
- Do not self-approve semantic validity without supervisor evidence.