long-horizon-prompting
This skill should be used when writing, enhancing, or evaluating the launch prompt for a long-running autonomous agent or a parallel multi-agent orchestration attacking a hard problem: pseudo-formal task briefs that define terms and an exact success predicate linguistically, enumerate non-counting outcomes, set persistence rules with explicit stop and return conditions and effort floors, manage a diverse portfolio of parallel approaches with an approach registry and blocked-route bookkeeping, and gate the return on adversarial audit. Route agent topology and coordination protocols to multi-age
npx skills add guanyang/open-agent-hub --skill long-horizon-prompting --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Directs how to craft the launch prompt for a long-running autonomous agent or parallel multi-agent system. It emphasizes a pseudo-formal task brief structure that defines terms, an exact success predicate, non-counting outcomes, persistence and stop/return rules, and a registry and bookkeeping for multiple parallel approaches. It also instructs routing topology and coordination to appropriate pattern and governance surfaces to support evaluation and auditing.
How it works
The skill prescribes writing a long-horizon brief with four components: definitions with degenerate cases, an exact success predicate with quantified scope, a enumerated list of non-counting outcomes, and enumerated failure modes for the auditor. It then covers orchestration policy for parallel workers via a registry of approach families and blocked-route bookkeeping, and a reporting contract with concrete artifacts and an auditable return condition. It also enforces persistence mechanisms (effort floors, stop conditions, return gates) and a contamination guard on external search. The workflow includes setting up a return condition that survives adversarial audit and a lean prompt structure that foregrounds the success predicate, non-counting list, and failure modes.
When to use it
Activate when writing or evaluating the launch prompt for a long-running autonomous run, converting vague problems into explicit briefs, designing root/orchestrator prompts for many parallel workers, adding persistence instructions and return gates, diagnosing failed runs due to brief issues, or building a pre-launch review step to enhance prompts before expensive agent time is spent.
What it can touch
The skill references a multi-agent pattern and governance surfaces, an approach registry, and blocked-route bookkeeping as tooling to support orchestration. It also mentions an adversarial auditor and a concrete artifact reporting contract, implying interaction with evaluation tooling and audit processes.
Caveats
It cautions that persistence pressure can increase the risk of cheating if not paired with verification gates, and emphasizes modular, independently checkable output to aid verification. It notes that the brief should be lean and outcome-first, with explicit failure modes for auditors. Licensing is MIT for the skill, and it relies on Claude Code as a declared tool.
# Long-Horizon Prompting 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. ## When to Activate Activate this skill when: - Writing or reviewing the prompt for a long-run
- When to Activate
- Core Concepts
- Pseudo-Formal Task Specification
- Anatomy of a Long-Horizon Brief
- Persistence Cuts Both Ways
- The Verification Bottleneck
- Structural Diversity in Parallel Search
- Stop Conditions, Effort, and Progress State
- Lean and Outcome-First
- Detailed Topics
- The CDC Prompt, Dissected
- Vendor Doctrine
- Generalizing Beyond Mathematics
- Practical Guidance
What does the long-horizon-prompting skill do?
This skill should be used when writing, enhancing, or evaluating the launch prompt for a long-running autonomous agent or a parallel multi-agent orchestration attacking a hard problem: pseudo-formal task briefs that define terms and an exact success predicate linguistically, enumerate non-counting outcomes, set persistence rules with explicit stop and return conditions and effort floors, manage a diverse portfolio of parallel approaches with an approach registry and blocked-route bookkeeping, and gate the return on adversarial audit. Route agent topology and coordination protocols to multi-age
How do I install it?
Run `npx skills add guanyang/open-agent-hub --skill long-horizon-prompting --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From guanyang/open-agent-hub, a repository with 940 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.