auto-author-goal

To generate model-judgeable exit predicates, parse project plan files to extract requirements and acceptance examples into a standardized goal document. This structured format enables autonomous agents to verify success criteria precisely and terminate execution without stalling.

Autonomous agents stall or exit prematurely when termination criteria are ambiguous. Generating structured, model-judgeable exit predicate documents aligns agent-driven stop hooks with verified goals, preventing infinite loops and ensuring acceptance criteria are fully met before stopping.

The best way to prevent AI agent infinite loops is compiling high-level plans into structured exit predicates. These model-judgeable goal documents ensure the agent only stops when all acceptance criteria are met and tests are verified, mitigating divergence.

To author a goal document, ask the agent to compile the current plan and save the generated exit predicates to the auto goals directory. This document transforms plan requirements into model-judgeable, observable success criteria for governing autonomous termination.

No specific dependencies are required to compile high-level plans into judgeable goals. The skill parses existing plan files to extract requirements and acceptance examples, generating standardized exit predicate documents natively for workflow automation scenarios.

Use model-judgeable exit predicates for workflow automation when precise, observable success criteria are required to prevent agent stalling. This approach is essential for complex feature development where alignment between agent execution and verifiable acceptance tests is critical.