菜单
Use this skill when the task is to reproduce or analyze a concrete AutoLabOS TUI/web symptom, especially when stale state, fresh-vs-existing divergence, resume mismatch, or persisted-artifact-vs-UI disagreement may be involved.
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
| name | tui-state-validation |
| description | Use this skill when the task is to reproduce or analyze a concrete AutoLabOS TUI/web symptom, especially when stale state, fresh-vs-existing divergence, resume mismatch, or persisted-artifact-vs-UI disagreement may be involved. |
| contract_version | 1 |
| contract_kind | codex_skill |
| runtime_contract | true |
| gate | live_state_reproduction_before_fix |
| validation | tui_state_validation_record |
Produce a structured live-validation result for one concrete target before proposing or evaluating a code change.
Treat actual user-visible behavior as the source of truth and compare, when relevant:
This skill is for narrowing the problem to the most likely minimal failing boundary before patching.
Use this skill when the user asks to:
Typical trigger phrases:
Always produce these sections:
/doctor output and current runtime context first when applicable.persisted_state_bugin_memory_projection_bugrefresh_render_bugresume_reload_bugrace_timing_bugUse when Codex should operate AutoLabOS as a research governance layer: create governed briefs, audit run artifacts, review paper readiness, strengthen weak nodes, or package traceable research bundles.
Use this skill when preparing, cleaning, or rerunning an AutoLabOS live validation workspace, especially when briefs, templates, run outputs, nested validation folders, or stale artifacts may affect a real paper-readiness run.
Use when editing AutoLabOS public source, tests, examples, or generated-code fixtures to prevent one-off experiment identifiers, model names, benchmark names, or condition markers from becoming public contracts.
Use this skill when validating AutoLabOS-generated TeX/PDF/paper bundles and keeping test/output coherent without manually substituting artifacts.
Use this skill when the goal is to move AutoLabOS outputs from workflow completion toward evidence-backed, baseline-bearing, paper-scale research quality.
Use this skill when the task is to repeatedly run a real TUI validation loop from test/ itself as the execution environment, reproduce issues, record them in ISSUES.md, apply the smallest plausible fix, revalidate the same flow, and keep validation grounded in node-owned execution rather than external substitution.