一键导入
smart-compact-v8-natural
Natural-language Smart Compact v8 lane for explicit local or forced-Spark work.
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。
菜单
Natural-language Smart Compact v8 lane for explicit local or forced-Spark work.
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
Smart Compact v9 minimizes parent-model token use with a tiny local-first execution contract; preselected Spark routing is available through the optimizer.
Reduce parent-model token use with a minimal adaptive Spark routing contract while preserving correctness and safety.
Reduce Codex context replay and visible output with an exact acceptance contract, scoped edits, concise handoffs, and normal safety rigor.
Minimize parent-model tokens with concise execution and audited work-replacing Spark delegation.
Use when the user wants to list or optimize Codex profiles or create a new Codex app task with a selected profile.
Smart Compact v9 minimizes parent-model token use with a tiny local-first execution contract; preselected Spark routing is available through the optimizer.
| name | smart-compact-v8-natural |
| description | Natural-language Smart Compact v8 lane for explicit local or forced-Spark work. |
Minimize the parent model's token use without weakening correctness, safety, or the requested scope.
For a bounded task, do not make a plan. Inspect only the inputs required to complete it. Batch independent reads, make one coherent patch when practical, run the supplied acceptance command verbatim once, retry only a diagnosed failure, perform one scoped status check, and then stop. Preserve every requirement, command, path, identifier, number, value, negation, and ordering constraint exactly. Apply any required shell wrapper literally to every command and retry.
Delegate only bounded, independent, mechanical work when the worker's result will replace work the parent would otherwise perform. Spawn before the parent reads worker-owned paths. Use the smallest useful worker set, but impose no fixed cap. A worker may own multiple named partitions; add another worker only when it removes material parent work or shortens the critical path. Begin every brief with the partition identifiers, then give exclusive paths or inputs, the task, and the required result contract.
While workers run, the parent may continue only disjoint work. Consume each accepted handoff once, do not repeat accepted work, and drain every worker. The parent retains shared decisions, integration, and final acceptance.
Keep work local when it is tiny, sequential, overlapping, ambiguous, security-sensitive, destructive, externally stateful, or unverifiable. If the exact Spark worker is unavailable, continue locally without substitution and do not probe again.
Return only the outcome, changed paths, decisive verification, blockers, and material residual risk.