菜单
Use when resuming an existing v1 bounded_runtime goal with goal.control.json and runtime.control.json.
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
Use when a formed task needs durable human-approved meaning, completion obligations, evidence claims, non-goals, or live/destructive authority boundaries.
Use when maintaining or resuming an existing v1 requirements.control.json run; new runs use forming-cybernetic-commitments.
Use when an existing v1 run must compile its reviewed control chain into runtime.control.json.
Use when optional local workflow guidance or legacy v1 workflow compatibility needs diagnosis.
Use when an existing v1 control chain requires design.control.json; v2 strategies remain agent-owned.
Use when cybernetic routing is requested but user input is pre-task intent or role-ambiguous context: confusion, dissatisfaction, risk sense, observed symptoms, failed attempts, method preference, process distrust, source material, declared current state, or unclear requested transformation.
| name | using-bounded-control-json |
| description | Use when resuming an existing v1 bounded_runtime goal with goal.control.json and runtime.control.json. |
Legacy v1 compatibility only. Do not start new bounded_runtime runs.
Use this skill for bounded_runtime goals. It is intentionally not the
JSON pre-goal control-chain executor.
bounded_runtime keeps its meaning only if it avoids the expanded
requirements/design/goal/plan/review/runtime chain. If execution needs
design, execution policy, semantic quality review, required-outcome coverage,
subagent coordination, or multi-stage control decisions, stop and route the
work to a controlled run instead.
Bounded runtime reads only:
goal.control.jsonruntime.control.jsonIt does not require requirements/design/plan/review control JSON.
Runtime writes only:
progress.jsonlruntime-status.jsonfinal-report.jsonDo not rewrite goal.control.json or runtime.control.json during runtime.
Before executing target work, run:
python3 .agents/skills/using-bounded-control-json/scripts/validate_bounded_runtime.py docs/cybernetics/runs/<slug>
Stop if validation fails. Report the missing bounded field or the smallest human decision needed to continue.
Append runtime observations to progress.jsonl as JSONL events. A required
step is complete only when a passing mainline step.completed event names that
step and includes every evidence id listed in runtime.control.json.
final-report.json records the bounded runtime completion claim and evidence.
It does not grant verifier permission to itself.
Before accepting or reporting goal_achieved: true, run:
python3 .agents/skills/using-bounded-control-json/scripts/verify_bounded_runtime.py docs/cybernetics/runs/<slug>
Treat this verifier as a structural final-claim gate: it checks that the bounded runtime claim matches the bounded JSON and recorded evidence. It is not a semantic quality gate. If the task needs adversarial semantic review, counterexamples, source-requirement preservation, or independent challenge, stop and route it to a controlled-run Counterexample Gate instead of accepting bounded verifier success as quality proof.
Verifier output reports semantic_quality_review: not_performed and lists
not_valid_for cases. Do not suppress or reinterpret those fields as quality
approval.
/goal AdapterThe /goal entry is a short pointer, not a control fact:
/goal Use .agents/skills/using-bounded-control-json and execute docs/cybernetics/runs/<slug>/runtime.control.json. Validate the bounded runtime first. Treat goal.control.json and runtime.control.json as read-only. Write only progress.jsonl, runtime-status.json, and final-report.json. If the bounded JSON is missing, invalid, inconsistent, or insufficient, stop and report the smallest required human decision.