What problem does it solve? Long-running autonomous agents and parallel multi-agent orchestrations fail when their launch prompts are vague: agents return answer-shaped near misses, converge prematurely on one approach, fabricate progress, or quit early. This Skill provides a structured method for writing launch briefs that specify exactly what counts as success and what does not. ## Core Features & Use Cases - Pseudo-Formal Brief Anatomy: Defines the ten blocks of a long-horizon brief, including definitions with degenerate cases, an exact success predicate, non-counting outcomes, orchestration heuristics, adversarial verification, and audit-gated return conditions. - Pre-Launch Evaluation Rubric: Scores any brief against ten adversarial-review dimensions before committing expensive agent compute, plus a red-team pass to find loopholes. - Reusable Template and Evidence Base: Ships a fill-in task brief template, an annotated real-world exemplar (the Cycle Double Cover orchestration prompt), dated vendor doctrine from OpenAI and Anthropic, and the academic research backing each design element. - Use Case: Before launching a 64-agent orchestration to attack an open research problem, convert the vague goal into a brief with a checkable success predicate, an enumerated list of near misses that do not count, and an auditor checklist of domain-specific failure modes. ## Quick Start Ask the agent to convert your hard problem statement into a pseudo-formal long-horizon task brief with a success predicate, non-counting outcomes, and an adversarial audit checklist before launching the run.
