菜单
Visible Harness prepare entry. Use after Grill readiness is approved to acquire datasets, clone baselines, run baseline smoke checks, and complete prepare through workflow-supervisor.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Internal Harness instruction source for auto-iterate-goal. Route through visible Harness aliases or hook contracts instead of invoking directly.
Internal Harness instruction source for auto-paper. Route through visible Harness aliases or hook contracts instead of invoking directly.
Internal Harness instruction source for evaluate. Route through visible Harness aliases or hook contracts instead of invoking directly.
Internal Harness instruction source for iterate. Route through visible Harness aliases or hook contracts instead of invoking directly.
Internal Harness instruction source for orchestrator. Route through visible Harness aliases or hook contracts instead of invoking directly.
Visible Harness write entry. Use for manuscript writing, citation-supported paper work, final documentation, README hardening, and GitHub Pages preparation.
| name | prepare |
| description | Visible Harness prepare entry. Use after Grill readiness is approved to acquire datasets, clone baselines, run baseline smoke checks, and complete prepare through workflow-supervisor. |
| argument-hint | [--complete] |
| allowed-tools | Read, Bash, Glob, Grep |
Use /prepare as the human-facing alias for the Execution Supervisor
prepare action. This is not a separate Skill Contract.
Read and follow:
../../../.claude/skills/workflow-supervisor/SKILL.md../../../.claude/shared/workflow-supervisor-runtime.md../../../CLAUDE.md../../../AGENTS.mdDefault command:
python tooling/workflow_supervisor/scripts/workflow_ctl.py start --segment prepare --complete --json
Before the default command, inspect Grill artifacts and readiness for a
Model Weight Ledger, model_weight_* policy rows, hf_model_access_policy,
and target_paths entries such as model_cache or model_<id>. Pull or
verify only candidate/approved weights with explicit source, access policy, and
target: exact local path, Hugging Face model id/repo via huggingface-cli download or git clone when available, or official checkpoint/release URL.
Use models/<id> only when no target is specified and the policy allows it.
Do not guess, download deferred/rejected/approval-required weights, or bypass
gated access. Report weight acquisition Gate ledger entries and artifacts
alongside the prepare command result.