Summary

This skill provides architectural guidance for building continuous autonomous agent loops with quality gates and recovery controls. It documents a decision tree for selecting loop patterns (sequential, PR-based, RFC-decomposed, or exploratory), references a recommended production stack combining RFC pipelines with quality gates and eval harnesses, and outlines failure modes and recovery procedures for common loop failure scenarios.

Detected Capabilities

Trigger Keywords

Phrases that agents use to match this skill to user intent.

Use Cases

• {Implement a self-checking autonomous agent loop that validates its own outputs before proceeding}
• Design a multi-stage agent loop combining RFC decomposition, code quality gates, and evaluation harnesses
• Diagnose and recover from agent loop failures like churn, repeated retries, or merge queue stalls
• Select the appropriate agent loop pattern based on control requirements (CI/PR, exploratory, or default sequential)

Quality Notes

• Skill provides high-level architectural guidance without prescriptive step-by-step instructions
• References external tools and skills (ralphinho-rfc-pipeline, plankton-code-quality, eval-harness, nanoclaw-repl) without documentation or links
• Loop selection flowchart is clear and helpful for decision-making
• Failure modes are relevant and practical, but recovery procedures lack detailed steps
• Skill assumes familiarity with referenced tools and concepts (RFC pipelines, quality gates, eval harnesses)
• No concrete code examples or implementation details provided
• License file is present but not referenced in content
• Missing: specific criteria for when to use each loop pattern, implementation examples, or integration instructions