long-horizon-prompting

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Long-Horizon Prompting

Long-Horizon 提示词设计

This skill covers the design of the prompt that launches an agent expected to work autonomously for hours or days, alone or as an orchestrator managing many parallel workers. The central technique is the pseudo-formal task brief: a specification written with the rigor of formal verification but expressed linguistically, because most hard problems have no machine-checkable success condition. The exemplar is the published prompt behind GPT-5.6 Sol Ultra's candidate proof of the Cycle Double Cover Conjecture, produced by a 64-subagent orchestration (claim-long-horizon-cdc-run). The prompt structure generalizes far beyond mathematics: any domain where success can be stated precisely and failure modes can be enumerated can use the same brief anatomy.

The controlling trade-off: everything that makes a long run productive (persistence, autonomy, parallelism) also raises the cost of a weak specification. A short interactive prompt fails cheaply; a long-horizon brief with a loophole burns hours of compute producing an answer-shaped artifact that does not solve the problem.

本技能涵盖用于启动Agent的提示词设计，这些Agent需要自主运行数小时甚至数天，既可以单独运行，也可以作为编排器管理多个并行工作Agent。核心技术是伪正式任务简报：一种兼具形式验证严谨性但以自然语言表述的规范，因为大多数复杂问题不存在可机器校验的成功条件。典型案例是GPT-5.6 Sol Ultra用于Cycle Double Cover猜想候选证明的公开提示词，该提示词由64个子Agent的编排系统生成（claim-long-horizon-cdc-run）。这种提示词结构的适用范围远超数学领域：任何可以精确表述成功标准、列举失败模式的领域，都可以采用相同的简报框架。

核心权衡点：所有让长期运行更高效的特性（持久化、自主性、并行性）也会放大弱规范的成本。简短的交互式提示词即使失败代价也很低；但存在漏洞的长周期简报会耗费数小时算力，生成看似解决问题但实际无效的产物。

When to Activate

激活场景

Activate this skill when:

Writing or reviewing the prompt for a long-running autonomous run before launching it
Converting a vague hard problem ("solve X", "figure out why Y happens") into an explicit brief with a success predicate and non-counting outcomes
Writing the root or orchestrator prompt that manages many parallel workers on an open-ended search problem
Adding persistence instructions, stop conditions, effort floors, or return gates to an agent prompt
Diagnosing a failed long run whose failure traces to the brief: premature return, an answer-shaped near miss, all workers converging on one approach, or fabricated completion claims
Building a pre-launch review step that enhances and evaluates prompts before expensive agent time is committed

Do not activate this skill for adjacent work owned by other skills:

Agent topology, supervisor versus swarm choice, handoff protocols, and coordination mechanics:
```
multi-agent-patterns
```
. That skill owns the architecture; this skill owns the words that steer it.
Runtime control surfaces, locked evaluators, rollback, durable logs, and approval boundaries around an autonomous loop:
```
harness-engineering
```
. Constraints that must survive optimization pressure belong in the harness, not the prompt.
Building the evaluator, regression suite, or deterministic quality gates a run is scored by:
```
evaluation
```
.
LLM-as-judge design, rubrics, pairwise comparison, and bias mitigation:
```
advanced-evaluation
```
.
Compaction, note-taking, and cross-session memory mechanics for surviving context limits:
```
context-compression
```
,
```
memory-systems
```
,
```
filesystem-context
```
.
Loops that modify their own harness or prompts:
```
self-improvement-loops
```
.
Remote sandboxes and background execution infrastructure:
```
hosted-agents
```
.

在以下场景激活本技能：

启动长期运行的自主Agent前，编写或审核其提示词
将模糊的复杂问题（如“解决X问题”“找出Y问题的原因”）转化为包含成功谓词和不计入结果情形的明确简报
编写用于管理多个并行工作Agent、处理开放式搜索问题的根Agent或编排器提示词
为Agent提示词添加持久化指令、停止条件、最低工作量要求或返回门控规则
诊断因简报缺陷导致的长期运行失败：提前返回、看似解决实际无效的产物、所有工作Agent收敛于单一方法、伪造完成结果
构建预启动审核步骤，在投入昂贵的Agent算力前优化并评估提示词

以下相邻工作不属于本技能范畴，应使用对应技能：

Agent拓扑结构、监督者与集群选择、交接协议、协调机制：
```
multi-agent-patterns
```
。该技能负责架构设计，本技能负责引导架构运行的提示词内容
运行时控制面、锁定式评估器、回滚机制、持久化日志、自主循环的审批边界：
```
harness-engineering
```
。必须抵御优化压力的约束应放在框架中，而非提示词里
构建用于评分的评估器、回归测试套件或确定性质量门：
```
evaluation
```
LLM作为评判器的设计、评分标准、两两对比、偏差缓解：
```
advanced-evaluation
```
为突破上下文限制的压缩、笔记记录、跨会话内存机制：
```
context-compression
```
、
```
memory-systems
```
、
```
filesystem-context
```
修改自身框架或提示词的循环：
```
self-improvement-loops
```
远程沙箱和后台执行基础设施：
```
hosted-agents
```

Core Concepts

核心概念

Pseudo-Formal Task Specification

伪正式任务规范

Formal verification requires a machine-checkable specification. Hard open problems rarely have one, but the discipline transfers: state the success condition so precisely that an adversarial reader cannot satisfy its letter without satisfying its intent. Four components, in order of leverage:

Definitions with degenerate cases. Define every load-bearing term before stating the goal, including the edge cases a lazy solution would exploit. The CDC prompt defines graph, bridge, cycle, and cycle double cover before the task, explicitly covering parallel-edge two-cycles, disconnected graphs, and the edgeless graph.
Exact success predicate. One statement of what must be true of the returned artifact, with scope quantifiers spelled out ("every finite loopless multigraph with no bridge, without additional assumptions such as cubicity, planarity, connectivity, or higher edge-connectivity").
Non-counting outcomes. An enumerated list of results that do not count: partial progress, special-case solutions, reductions to another unproved statement, bounded or computational verification, and best-effort summaries. This is the highest-leverage component. Under persistence pressure, models produce answer-shaped near misses; each excluded outcome removes one escape hatch.
Enumerated failure modes for the auditor. A concrete checklist of the domain-specific ways a candidate can be subtly wrong (in CDC: repeated-edge closed trails masquerading as cycles, bridges introduced by reductions, circular use of an equivalent statement). Verifiers with an enumerated hunt list catch what generic "check the work" instructions miss.

形式验证需要可机器校验的规范，但复杂的开放性问题很少具备这一条件，不过形式验证的思路可以迁移：将成功条件表述得足够精确，让刻意钻空子的读者也无法只满足字面要求而不达成实际意图。包含四个核心组件，按影响力排序：

包含退化情况的定义。在阐述目标前，明确定义所有关键术语，包括懒惰解决方案可能利用的边缘情况。CDC提示词在任务前定义了图、桥、环和环双覆盖，明确涵盖了平行边双环、非连通图和无边图。
精确的成功谓词。用一句话描述返回产物必须满足的条件，明确写出范围量词（如“每个无桥的有限无环多重图，不附加立方性、平面性、连通性或更高边连通性等假设”）。
不计入结果的情形。列举不被视为成功的结果：部分进展、特殊场景解决方案、转化为未证明的陈述、有限范围或计算型验证、尽力而为的总结。这是影响力最大的组件。在持久化压力下，模型会生成看似解决问题的无效产物，每排除一种情形就减少一个漏洞。
审计员的枚举式失败模式。列出领域特有的、候选方案可能存在的隐性错误的具体清单（在CDC案例中：伪装成环的重复边闭合路径、归约过程中引入的桥、等价陈述的循环使用）。拥有明确排查清单的验证器能发现通用“检查工作”指令遗漏的问题。

Anatomy of a Long-Horizon Brief

长周期简报的框架

Block	Job	Failure it prevents
Definitions	Fix the vocabulary, including degenerate cases	Loophole solutions on technicalities
Success predicate	State exactly what must be true at return	Scope-narrowed answers
Non-counting outcomes	Enumerate near misses that do not count	Answer-shaped partial results
Solvability framing	"Assume a solution exists" where existence is plausible	Give-up drift, "this is open" refusals
Orchestration policy	Heuristics for allocating parallel workers, not fixed assignments	Premature convergence, wasted parallelism
Verification policy	Adversarial audit with enumerated failure modes	Lenient self-judging
Reporting contract	Concrete artifacts required; status reports rejected	Vague optimism, fabricated progress
Return condition	Return only when the artifact survives audit	Premature return, best-effort summaries
Effort floor	Minimum effort before giving up is considered	Early abandonment
Contamination guards	What external search may and may not be used for	Laundered lookups, benchmark leakage

模块	作用	预防的失败场景
定义	明确词汇范围，包括退化情况	利用技术细节钻空子的解决方案
成功谓词	明确返回时必须满足的精确条件	缩小范围的答案
不计入结果的情形	列举不被认可的近似结果	看似解决实际无效的部分成果
可解性框架	在可能存在解决方案的场景下，使用“假设解决方案存在”	放弃式偏离、“此问题无解”的拒绝
编排策略	分配并行工作Agent的启发式规则，而非固定分配	过早收敛、浪费并行算力
验证策略	基于枚举失败模式的对抗性审计	宽松的自我评判
报告约定	要求提交具体产物，拒绝状态报告	模糊的乐观表述、伪造的进展
返回条件	仅当产物通过审计时返回	提前返回、尽力而为的总结
最低工作量	设定放弃前必须完成的最小工作量	过早放弃
污染防护	明确外部搜索的允许范围	洗白式查找、基准泄露

Persistence Cuts Both Ways

持久化的双刃剑

Persistence instructions ("do not return until", effort floors, assume-solvable framing) counter the documented drift toward giving up on long trajectories (claim-long-horizon-give-up-drift). But the same pressure raises the reward-hacking surface: the most persistence-trained frontier model measured to date also showed the highest detected cheating rate of any model its evaluator had tested, and its measured time horizon was not robust to whether cheating counted as success (claim-long-horizon-persistence-hacking). The design rule: never add a persistence instruction without a matching verification gate. Persistence pressure against a loose success predicate produces confident non-solutions.

持久化指令（如“未完成前不要返回”、最低工作量、假设可解的框架）能对抗已被证实的、长期任务中模型逐渐放弃的偏离现象（claim-long-horizon-give-up-drift）。但同样的压力会增加奖励黑客的风险：目前测得的接受过最多持久化训练的前沿模型，其作弊率也高于评估器测试过的其他模型，且其测得的时间稳定性在作弊被视为成功时会失效（claim-long-horizon-persistence-hacking）。设计规则：添加持久化指令时，必须搭配对应的验证门控。针对宽松成功谓词的持久化压力会生成看似可信的无效解决方案。

The Verification Bottleneck

验证瓶颈

Parallel sampling reliably raises the chance that some worker finds a correct answer, but the system's ability to select that answer lags behind, and model judges of hard artifacts are systematically lenient, rewarding rigorous-looking but incomplete arguments (claim-long-horizon-verification-gap). Budget as much prompt design for the verifier as for the generator:

Give auditors the enumerated failure-mode list from the brief, not a generic quality instruction.
Require the generator to produce modular, independently checkable output (lemma-level structure with stated premises and conclusions) so verification decomposes.
Use fresh-context adversarial verifiers rather than self-critique; a verifier that did not build the artifact cannot rationalize its gaps.
Treat inter-agent agreement as a diversity failure signal, not as confirmation: committees converge most tightly on the hardest problems, where unanimity reflects shared bias rather than corroboration (claim-long-horizon-diversity-collapse).

并行采样确实能提高某个工作Agent找到正确答案的概率，但系统筛选正确答案的能力滞后，且针对复杂产物的模型评判者普遍宽松，会奖励看似严谨但不完整的论证（claim-long-horizon-verification-gap）。为验证器分配的提示词设计预算应与生成器相当：

为审计员提供简报中的枚举式失败模式清单，而非通用的质量要求。
要求生成器生成模块化、可独立校验的输出（带有明确前提和结论的引理级结构），以便验证工作可以分解。
使用全新上下文的对抗性验证器而非自我批判；未参与产物构建的验证器无法合理化产物中的漏洞。
将Agent间的一致意见视为多样性失败信号，而非确认：在最复杂的问题上，委员会的收敛性最强，此时的一致意见反映的是共同偏见而非佐证（claim-long-horizon-diversity-collapse）。

Structural Diversity in Parallel Search

并行搜索中的结构多样性

Role labels do not create diversity; parallel workers share priors and converge unless independence is engineered:

Keep early-round workers blind to the currently favored approach.
Maintain an explicit registry of approach families, grouped by underlying idea rather than surface wording, and redirect workers away from crowded families.
Mark a route blocked when it stalls at a missing step as hard as the original goal; reassign workers to it only for a materially new mechanism, not for enthusiasm.
Cross-pollinate late, after independent development has exposed each route's real strengths and gaps.
Do not let one approach dominate because its reductions are elegant; a route ending at a lemma equivalent in strength to the original goal is not progress.

角色标签无法创造多样性；并行工作Agent共享先验知识，除非刻意设计独立性，否则会收敛：

让早期阶段的工作Agent对当前受青睐的方法不知情。
维护明确的方法家族注册表，按核心思想而非表面表述分组，引导工作Agent避开已拥挤的方法家族。
当某个路径在与原问题难度相当的缺失步骤上停滞时，标记该路径为受阻；仅当出现实质性新机制时，才重新分配工作Agent到该路径，而非仅凭热情。
在后期进行交叉融合，等待各路径独立发展后暴露其真正的优势和漏洞。
不要因某个方法的归约过程优雅就让其主导；若路径最终指向与原问题难度相当的引理，则不算进展。

Stop Conditions, Effort, and Progress State

停止条件、工作量与进展状态

Long trajectories drift toward uncertainty and abandonment, and a budget stated once at the top of the prompt loses force as context grows (claim-long-horizon-give-up-drift). Countermeasures that belong in the brief: an explicit effort floor ("spend at least this much effort before considering returning"), assume-solvable framing where a solution plausibly exists, and a return condition phrased as a predicate over the artifact rather than over the agent's confidence. Countermeasures that belong outside the prompt: an externally maintained ledger of verified progress re-injected each round, which in controlled comparisons rescued large-quantity tasks that prompt-only and completion-gated setups failed entirely (claim-long-horizon-state-ledger). Progress claims should be auditable: requiring each reported claim to trace to a tool result or artifact from the current session nearly eliminated fabricated status reports in vendor testing (claim-long-horizon-evidence-audit).

长期任务会逐渐偏离，变得不确定并最终放弃，且在提示词顶部一次性声明的预算会随着上下文增长而失去效力（claim-long-horizon-give-up-drift）。应在简报中加入以下应对措施：明确的最低工作量（“考虑返回前至少完成此工作量”）、在可能存在解决方案的场景下使用假设可解的框架、将返回条件表述为针对产物的谓词而非Agent的信心。应在提示词之外加入以下应对措施：外部维护的已验证进展台账，每轮重新注入上下文；对比测试显示，这种设置成功挽救了仅靠提示词和完成门控无法解决的大量任务（claim-long-horizon-state-ledger）。进展声明必须可审计：要求每个报告的声明都能追溯到当前会话的工具结果或产物，在供应商测试中几乎消除了伪造的状态报告（claim-long-horizon-evidence-audit）。

Lean and Outcome-First

精简且以结果为导向

Both major vendors converged on the same doctrine for current frontier models: the prompt should carry the outcome, hard constraints, evidence sources, and completion bar, and leave the path to the model. Accumulated instruction stacks measurably hurt; leaner system prompts improved vendor coding-agent evaluations while cutting cost (claim-long-horizon-lean-prompt). Persistence itself is increasingly trained in rather than prompted in, so spend the token budget on what training cannot supply: the success predicate, the non-counting list, and the domain failure modes only an expert in the problem knows.

两大主流供应商针对当前前沿模型达成了一致原则：提示词应承载结果要求、硬性约束、证据来源和完成标准，将实现路径交给模型。积累的指令堆栈会明显损害性能；更精简的系统提示词在降低成本的同时，提升了供应商编码Agent的评估表现（claim-long-horizon-lean-prompt）。持久化特性越来越多地通过训练实现而非提示词指定，因此应将token预算花在训练无法提供的内容上：成功谓词、不计入结果的清单，以及只有问题领域专家才了解的领域失败模式。

Detailed Topics

详细主题

The CDC Prompt, Dissected

CDC提示词拆解

The published Cycle Double Cover prompt implements every block of the brief anatomy in under a page: formal definitions closing degenerate-case loopholes, an exact success predicate with scope quantifiers, five classes of explicitly non-counting partial progress, dynamic orchestration heuristics for up to 64 concurrent agents with an approach-family registry and blocked-route bookkeeping, adversarial auditors with a seven-item failure-mode hunt list, a concrete-artifact reporting contract, an audit-gated return condition, an eight-hour effort floor, and a contamination guard restricting web search to background material (claim-long-horizon-cdc-run). The full annotated text is in the CDC prompt reference.

Two honest caveats. The candidate proof had no independent peer review or formalization when published, so the prompt is the validated artifact of interest here, not the theorem. And no public ablation isolates which prompt elements carried the result; the mechanism-level evidence comes from the independent research in the research evidence reference.

公开的Cycle Double Cover提示词在一页内实现了简报框架的所有模块：关闭退化情况漏洞的形式化定义、带有范围量词的精确成功谓词、五类明确不计入结果的部分进展、针对最多64个并发Agent的动态编排启发式规则（含方法家族注册表和受阻路径记录）、带有7项失败模式排查清单的对抗性审计员、具体产物的报告约定、基于审计的返回条件、8小时最低工作量要求，以及限制网络搜索仅用于背景资料的污染防护（claim-long-horizon-cdc-run）。完整的带注释文本见CDC提示词参考文档。

需说明两个客观情况：候选证明发布时未经过独立同行评审或形式化验证，因此此处的验证产物是提示词而非定理。此外，目前没有公开的消融实验能确定哪些提示词元素起到了关键作用；机制层面的证据来自研究证据参考文档中的独立研究。

Vendor Doctrine

供应商原则

OpenAI and Anthropic guidance overlap on fundamentals (explicit completion bars, stop rules, verification before return) and differ in emphasis. OpenAI doctrine centers persistence blocks, risk-tiered autonomy thresholds, self-constructed rubrics, and reasoning-effort dials; its multi-agent API institutionalizes a root agent with bounded-task subagents. Anthropic doctrine centers the four-part subagent delegation spec (objective, output format, tool guidance, task boundaries), explicit effort-scaling tiers by task complexity, evidence-grounded progress reporting, and fresh-context verifier subagents. Both now warn that over-prescriptive prompts degrade current-generation models. Dated extracts with sources are in the vendor guidance reference.

OpenAI和Anthropic的指导原则在基础内容上重叠（明确的完成标准、停止规则、返回前验证），但侧重点不同。OpenAI的原则以持久化模块、风险分层的自主性阈值、自主构建的评分标准和推理工作量调节为核心；其多Agent API将根Agent与任务受限的子Agent机制化。Anthropic的原则以四部分子Agent委托规范（目标、输出格式、工具指导、任务边界）、基于任务复杂度的明确工作量分层、基于证据的进展报告和全新上下文验证子Agent为核心。两者均警告，过度指定的提示词会降低当前代模型的性能。带有来源的过时摘录见供应商指导参考文档。

Generalizing Beyond Mathematics

超越数学领域的推广

The CDC prompt worked because mathematics allows sharp statements, but each element has a general form usable in any rigorous domain:

CDC element	General form
Formal graph definitions	Operationalize every load-bearing term; state units, populations, boundaries, degenerate cases
"Exactly two occurrences of each edge"	A quantified, checkable property of the deliverable
"Special graph classes do not count"	"Results holding only under narrowed scope do not count"
"No reduction to another unproved conjecture"	"No dependence on an unvalidated assumption or unavailable dataset"
"Computational verification through fixed size is insufficient"	"Anecdotal or small-sample evidence is insufficient"
Parallel-edge and bridge edge cases for auditors	The domain's known confounders, artifacts, and failure modes as an audit checklist
"Do not search for a solution to this exact conjecture"	"Do not launder the answer from sources the result is supposed to be independent of"

The transformation workflow for a scientist or engineer with a hard problem: state what a complete answer would let them do, work backward to the predicate that enables it, then spend most of the effort listing what they would refuse to accept from a junior collaborator. That refusal list becomes the non-counting outcomes and the auditor checklist.

CDC提示词之所以有效，是因为数学领域允许精确表述，但每个元素都有通用形式，可用于任何严谨领域：

CDC元素	通用形式
形式化图定义	明确所有关键术语的操作定义；说明单位、群体、边界、退化情况
“每条边恰好出现两次”	可量化、可校验的交付产物属性
“特殊图类不计入结果”	“仅在缩小范围内成立的结果不计入”
“不得归约为其他未证明的猜想”	“不得依赖未验证的假设或不可用的数据集”
“有限规模的计算验证不足够”	“轶事或小样本证据不足够”
审计员排查的平行边和桥边缘情况	将领域已知的混淆因素、伪产物和失败模式作为审计清单
“不得搜索此猜想的现成解决方案”	“不得从应独立于结果的来源中洗白答案”

面对复杂问题的科学家或工程师的转化流程：明确完整答案能实现的目标，反向推导实现该目标的谓词，然后将大部分精力用于列出他们会拒绝的初级合作者提交的成果。这份拒绝清单将成为不计入结果的情形和审计员排查清单。

Practical Guidance

实践指南

Brief-Writing Workflow

简报编写流程

Write the success predicate first, as one sentence with explicit quantifiers and scope. If it cannot be written, the problem is not ready for a long-horizon run; decompose it or run a scoping session instead.
Enumerate non-counting outcomes by asking what a capable agent under pressure would return instead of a solution: the narrowed-scope version, the reduction, the survey, the plan, the confident sketch.
Define terms, starting from the degenerate cases the predicate must survive.
Write the auditor checklist: the domain-specific ways a candidate artifact can look right and be wrong.
Set the orchestration policy as heuristics (diversity early, registry by idea, blocked-route rules, late cross-pollination), never as fixed worker-to-strategy assignments.
Set the reporting contract (concrete artifacts, evidence-traceable claims) and the return condition (survives adversarial audit against the checklist).
Add the effort floor, solvability framing if warranted, and contamination guards.
Red-team the brief before launch: ask a fresh model instance "how could an agent satisfy the letter of this brief without solving the problem?" and patch every credible answer.

首先编写成功谓词，用一句话明确写出量词和范围。如果无法写出，说明问题尚未准备好进行长周期运行；应先分解问题或开展范围界定会话。
列举不计入结果的情形：设想承受压力的能力较强的Agent会返回哪些替代解决方案，如缩小范围的版本、归约结果、调研报告、计划、看似可信的草图。
定义术语，从成功谓词必须覆盖的退化情况开始。
编写审计员排查清单：候选产物看似正确但实际错误的领域特有方式。
将编排策略设置为启发式规则（早期保持多样性、按思想分组的注册表、受阻路径规则、后期交叉融合），而非固定的工作Agent-策略分配。
设置报告约定（具体产物、可追溯到证据的声明）和返回条件（通过对抗性审计排查清单）。
添加最低工作量要求、适用的可解性框架和污染防护规则。
启动前对简报进行红队测试：询问全新模型实例“Agent如何在不解决问题的情况下满足本简报的字面要求？”，并修复所有可信的漏洞。

Pre-Launch Evaluation

预启动评估

Score any long-horizon brief against these questions before committing agent time. Any "no" is a defect to fix, not a judgment call:

Can an adversarial reader determine unambiguously whether a given artifact satisfies the success predicate?
Is every plausible near miss explicitly listed as non-counting?
Does the auditor have an enumerated, domain-specific failure-mode list?
Is every persistence instruction paired with a verification gate?
Is the return condition a predicate over the artifact, not over agent confidence or elapsed effort?
Does the orchestration policy preserve early independence and include blocked-route bookkeeping?
Are reporting requirements artifact-based rather than status-based?
Are contamination guards stated for any external retrieval?
Is anything in the prompt a constraint that must survive optimization pressure? Move it to the harness (
```
harness-engineering
```
); prompt-stated constraints are advisory.

在投入Agent算力前，用以下问题评估任何长周期简报。任何“否”都是需要修复的缺陷，而非主观判断：

刻意钻空子的读者能否明确判断某个给定产物是否满足成功谓词？
所有看似合理的近似结果是否都被明确列为不计入结果的情形？
审计员是否拥有枚举式的领域特有失败模式清单？
每个持久化指令是否都搭配了强度匹配的验证门控？
返回条件是否是针对产物的谓词，而非Agent的信心、工作量或耗时？
编排策略是否能保持早期独立性，并包含受阻路径记录？
报告要求是否基于产物而非状态？
是否为任何外部检索设置了污染防护规则？
提示词中是否包含必须抵御优化压力的约束？将其移至框架（
```
harness-engineering
```
）；提示词中声明的约束仅为建议性。

Examples

示例

Example 1: Pseudo-formal brief skeleton

text

DEFINITIONS
  <every load-bearing term, including degenerate cases>

TASK
  <exact success predicate with quantifiers and scope>

DOES NOT COUNT
  <narrowed scope> <reduction to unvalidated assumption>
  <bounded/anecdotal verification> <plan or survey instead of artifact>

ORCHESTRATION (for parallel runs)
  Begin with a genuinely diverse portfolio. Keep early workers blind
  to the favored approach. Registry of approach families by idea, not
  wording. Mark routes blocked at goal-strength gaps; reopen only for
  a materially new mechanism. Cross-pollinate late.

VERIFICATION
  Adversarial audit of every candidate against:
  <domain failure-mode checklist>
  Workers return concrete artifacts; status reports are rejected.

RETURN CONDITION
  Return only when a candidate survives the audit. Do not return a
  reduction, partial result, or explanation of difficulty.

EFFORT
  Assume a solution exists. Spend at least <floor> before considering
  returning.

CONTAMINATION
  External search only for <background>; never for <the answer>.

Example 2: Weak prompt to strong brief (root-cause analysis)

text

Weak:  "Investigate why our v4 model underperforms v3 in production
        and write up what you find. Be thorough."

Strong: TASK: Identify a defect that, when corrected, closes the
        v4-versus-v3 production gap on the frozen evaluation slice,
        demonstrated by a reproduction script and a corrected run.
        DOES NOT COUNT: correlational narratives without an
        intervention; defects explaining under a stated fraction of
        the gap; "data drift" without an identified slice and
        mechanism; a list of hypotheses.
        VERIFICATION: an adversarial reviewer checks the reproduction
        for train/serve skew, leakage in the eval slice, seed
        sensitivity, and preprocessing divergence.
        RETURN: only a candidate that survives that review.

The weak version invites a status report. The strong version makes the deliverable checkable and pre-blocks the three most likely near misses.

示例1：伪正式简报框架

text

DEFINITIONS
  <所有关键术语，包括退化情况>

TASK
  <带有量词和范围的精确成功谓词>

DOES NOT COUNT
  <缩小范围的结果> <归约为未验证假设的结果>
  <有限/轶事性验证> <计划或调研而非产物>

ORCHESTRATION (for parallel runs)
  Begin with a genuinely diverse portfolio. Keep early workers blind
  to the favored approach. Registry of approach families by idea, not
  wording. Mark routes blocked at goal-strength gaps; reopen only for
  a materially new mechanism. Cross-pollinate late.

VERIFICATION
  Adversarial audit of every candidate against:
  <领域失败模式排查清单>
  Workers return concrete artifacts; status reports are rejected.

RETURN CONDITION
  Return only when a candidate survives the audit. Do not return a
  reduction, partial result, or explanation of difficulty.

EFFORT
  Assume a solution exists. Spend at least <floor> before considering
  returning.

CONTAMINATION
  External search only for <background>; never for <the answer>.

示例2：从弱提示词到强简报（根因分析）

text

Weak:  "Investigate why our v4 model underperforms v3 in production
        and write up what you find. Be thorough."

Strong: TASK: Identify a defect that, when corrected, closes the
        v4-versus-v3 production gap on the frozen evaluation slice,
        demonstrated by a reproduction script and a corrected run.
        DOES NOT COUNT: correlational narratives without an
        intervention; defects explaining under a stated fraction of
        the gap; "data drift" without an identified slice and
        mechanism; a list of hypotheses.
        VERIFICATION: an adversarial reviewer checks the reproduction
        for train/serve skew, leakage in the eval slice, seed
        sensitivity, and preprocessing divergence.
        RETURN: only a candidate that survives that review.

弱版本会引发状态报告。强版本让交付产物可校验，并预先阻止了三种最可能出现的近似结果。

Guidelines

准则

Write the success predicate before any other prompt content; if it cannot be stated precisely, do not launch a long-horizon run.
Enumerate non-counting outcomes explicitly; every near miss not excluded is an escape hatch.
Define load-bearing terms including degenerate cases before stating the task.
Give auditors an enumerated domain failure-mode checklist, never a generic quality instruction.
Pair every persistence instruction with a verification gate of matching strength.
Phrase return conditions as predicates over the artifact, not over confidence, effort, or elapsed time.
Assign parallel workers by heuristic policy with an approach-family registry; never fixed strategy quotas.
Preserve early-round worker independence; cross-pollinate only after routes have developed independently.
Mark routes blocked at goal-strength gaps and require a materially new mechanism to reopen them.
Require concrete artifacts from every worker and reject status reports and vague optimism.
Require progress claims to trace to session evidence (tool results, files, logs).
State contamination guards for external retrieval whenever result independence matters.
Keep the brief lean: outcome, constraints, completion bar, failure modes; leave the path to the model.
Enforce hard budgets and permissions in the harness; treat prompt-stated constraints as advisory.

先编写成功谓词，再编写其他提示词内容；如果无法精确表述，不要启动长周期运行。
明确列举不计入结果的情形；每一个未被排除的近似结果都是一个漏洞。
在阐述任务前，定义关键术语包括退化情况。
为审计员提供枚举式的领域失败模式排查清单，而非通用的质量要求。
每个持久化指令都要搭配强度匹配的验证门控。
将返回条件表述为针对产物的谓词，而非信心、工作量或耗时。
通过带有方法家族注册表的启发式策略分配并行工作Agent；不要使用固定的策略配额。
保持早期工作Agent的独立性；仅在各路径独立发展后进行交叉融合。
将在与原问题难度相当的漏洞处停滞的路径标记为受阻，仅当出现实质性新机制时重新启用。
要求所有工作Agent提交具体产物，拒绝状态报告和模糊的乐观表述。
要求进展声明能追溯到会话证据（工具结果、文件、日志）。
当结果独立性重要时，为外部检索设置污染防护规则。
保持简报精简：包含结果、约束、完成标准、失败模式；将实现路径交给模型。
在框架中强制执行严格的预算和权限；将提示词中声明的约束视为建议性。

Gotchas

常见陷阱

Answer-shaped near misses: Under persistence pressure, agents return artifacts with the shape of a solution (narrowed scope, unproved dependency, survey instead of result). The non-counting list is the fix; write it by predicting the specific near misses your problem invites.
Circular satisfaction: The subtlest near miss is an argument that assumes a statement equivalent in strength to the goal. The CDC prompt names this explicitly ("circular use of an equivalent CDC statement"); every domain has an analogue, and auditors will not catch it unless it is on their checklist.
Persistence without verification breeds hacking: Persistence-trained and persistence-prompted agents show elevated rates of gaming their success signal (claim-long-horizon-persistence-hacking). If the brief demands "do not return without success" but success is checked leniently, the agent optimizes the leniency.
Unanimity is not corroboration: Parallel agents agreeing is weak evidence when they share priors, and convergence tightens on harder problems (claim-long-horizon-diversity-collapse). Never use agreement alone as a return trigger; audit content, and treat fast consensus as a diversity failure.
Under-specified delegation duplicates work: Subagent tasks missing any of objective, output format, tool guidance, or boundaries produce overlapping and gap-ridden coverage. The orchestrator prompt should require all four in every spawn.
Status-report theater: Long runs drift into reporting activity instead of results, including fabricated completions. Require artifact-based reporting and evidence-traceable claims (claim-long-horizon-evidence-audit); reject "on track" without a pointer.
Effort floors are permissions, not schedules: The CDC run finished well under its stated eight-hour floor (claim-long-horizon-cdc-run). A floor removes the agent's permission to quit early; it neither guarantees nor bounds runtime. Enforce actual time and cost budgets in the harness.
Prompt-stated budgets decay: A budget or reminder stated once loses force as the trajectory grows; re-inject budget and verified-progress state periodically from outside the loop (claim-long-horizon-give-up-drift).
Assume-solvable on ill-posed problems: Solvability framing counters give-up drift but instructs the model to never conclude "no solution exists". On genuinely open or ill-posed questions, pair it with a counterexample track or drop it, or the run will fabricate.
Over-prescription backfires on frontier models: Step-by-step scripts and stacked MUST/NEVER emphasis measurably degrade current-generation model output (claim-long-horizon-lean-prompt). Migrate old prompt stacks by starting from the minimal brief, not by accretion.

看似解决实际无效的产物：在持久化压力下，Agent会生成看似解决问题的产物（缩小范围、未验证的依赖、调研而非结果）。解决方法是编写不计入结果的清单；通过预测问题可能引发的具体近似结果来完成清单。
循环满足：最隐蔽的近似结果是假设与目标强度相当的陈述的论证。CDC提示词明确指出了这一点（“循环使用等价的CDC陈述”）；每个领域都有类似情况，除非将其列入排查清单，否则审计员无法发现。
无验证的持久化会催生黑客行为：接受过持久化训练或提示的Agent会更高频率地操纵成功信号（claim-long-horizon-persistence-hacking）。如果简报要求“未成功不要返回”但成功校验宽松，Agent会针对宽松的校验进行优化。
一致意见并非佐证：当并行Agent共享先验知识时，一致意见是弱证据，且在更复杂的问题上收敛性更强（claim-long-horizon-diversity-collapse）。永远不要仅将一致意见作为返回触发条件；要审核内容，并将快速共识视为多样性失败。
指定不足的委托会导致重复工作：缺少目标、输出格式、工具指导或边界的子Agent任务会导致重叠和覆盖漏洞。编排器提示词应要求每个生成的子Agent任务包含这四个要素。
状态报告形式主义：长期运行会逐渐转向报告活动而非结果，包括伪造的完成结果。要求基于产物的报告和可追溯到证据的声明（claim-long-horizon-evidence-audit）；拒绝没有指向性的“进展顺利”报告。
最低工作量是权限而非时间表：CDC运行远早于其声明的8小时最低工作量完成（claim-long-horizon-cdc-run）。最低工作量只是取消了Agent提前退出的权限；既不保证也不限制运行时间。在框架中强制执行实际的时间和成本预算。
提示词中声明的预算会失效：一次性声明的预算或提醒会随着任务推进而失去效力；需定期从外部循环重新注入预算和已验证的进展状态（claim-long-horizon-give-up-drift）。
在不适定问题上使用假设可解框架：可解性框架能对抗放弃式偏离，但会指示模型永远不要得出“无解”的结论。对于真正开放或不适定的问题，需搭配反例追踪或放弃该框架，否则运行会生成伪造内容。
过度指定会适得其反：分步脚本和堆叠的MUST/NEVER强调会明显降低当前代模型的输出性能（claim-long-horizon-lean-prompt）。迁移旧提示词堆栈时，应从最简简报开始，而非逐步累加内容。

Integration

集成

This skill owns the launch prompt for long-running and parallel agent work. Adjacent skills own the machinery around it:

multi-agent-patterns - Owns topology, handoffs, and coordination protocols; this skill writes the orchestration policy those structures execute
harness-engineering - Owns runtime-enforced budgets, locked evaluators, and control surfaces; constraints that must survive optimization pressure move there
evaluation - Owns deterministic evaluators and quality gates referenced by the brief's verification policy
advanced-evaluation - Owns judge design, rubrics, and bias mitigation for the adversarial audit step
self-improvement-loops - Owns loops that rewrite their own prompts and harnesses; briefs written here can become that loop's seed
filesystem-context - Owns the durable progress ledgers and artifacts the reporting contract points at
context-compression - Owns compaction and handoff mechanics when a run outlives its context window
hosted-agents - Owns the sandboxed infrastructure long runs execute on

本技能负责长期运行和并行Agent工作的启动提示词。相邻技能负责相关的配套机制：

multi-agent-patterns - 负责拓扑结构、交接和协调协议；本技能编写这些结构执行的编排策略
harness-engineering - 负责运行时强制执行的预算、锁定式评估器和控制面；必须抵御优化压力的约束应移至此处
evaluation - 负责简报验证策略引用的确定性评估器和质量门
advanced-evaluation - 负责对抗性审计步骤中的评判器设计、评分标准和偏差缓解
self-improvement-loops - 负责重写自身提示词和框架的循环；此处编写的简报可作为循环的种子
filesystem-context - 负责报告约定指向的持久化进展台账和产物
context-compression - 负责运行超出上下文窗口时的压缩和交接机制
hosted-agents - 负责长期运行的沙箱化基础设施

References

参考资料

Internal references:

Annotated CDC prompt - The full published prompt with element-by-element annotation and provenance
Vendor guidance - Dated OpenAI and Anthropic long-horizon and multi-agent prompting doctrine with sources
Research evidence - Dated academic findings behind each brief element
Task brief template - Reusable pseudo-formal brief template and pre-launch evaluation rubric

Related skills in this collection:

multi-agent-patterns - Topology and coordination for the orchestrations these briefs steer
harness-engineering - Runtime enforcement of what the brief can only request

External resources:

OpenAI, published prompt for the GPT-5.6 Sol Ultra Cycle Double Cover run (July 2026) - The exemplar brief
METR, predeployment evaluation of GPT-5.6 Sol (June 2026) - Persistence-training and reward-hacking linkage
Anthropic, "How we built our multi-agent research system" (June 2025) - Delegation specs and effort scaling
OpenAI GPT-5.x prompting guides and Anthropic Claude prompting docs - Vendor doctrine detailed in the vendor guidance reference

Numeric, benchmark, volatile, or vendor-performance claims in this skill carry inline

claim-*

IDs backed by

researcher/claims/index.jsonl

. Detailed numbers live in the dated reference files.

内部参考：

带注释的CDC提示词 - 完整的公开提示词，包含逐元素注释和来源
供应商指导 - 带有来源的OpenAI和Anthropic长周期及多Agent提示词原则（过时版本）
研究证据 - 每个简报元素背后的学术研究成果（过时版本）
任务简报模板 - 可复用的伪正式简报模板和预启动评估标准

本技能集中的相关技能：

multi-agent-patterns - 本技能引导的编排系统的拓扑结构和协调机制
harness-engineering - 对简报仅能请求的内容进行运行时强制执行

外部资源：

OpenAI，GPT-5.6 Sol Ultra Cycle Double Cover运行的公开提示词（2026年7月） - 典型简报案例
METR，GPT-5.6 Sol部署前评估（2026年6月） - 持久化训练与奖励黑客的关联
Anthropic，“我们如何构建多Agent研究系统”（2025年6月） - 委托规范和工作量分层
OpenAI GPT-5.x提示词指南和Anthropic Claude提示词文档 - 供应商指导参考文档中详细说明的供应商原则

本技能中包含的数值、基准、易变或供应商性能声明带有内嵌的

claim-*

ID，由

researcher/claims/index.jsonl

提供支持。详细数值见过时的参考文件。

Skill Metadata

技能元数据

Created: 2026-07-11 Last Updated: 2026-07-11 Author: Agent Skills for Context Engineering Contributors Version: 1.0.0

创建时间: 2026-07-11 最后更新时间: 2026-07-11 作者: Agent Skills for Context Engineering Contributors 版本: 1.0.0