Generate critical, contrarian, and steelman questions about repository design.
npx skills add https://github.com/anthony-chaudhary/fak --skill question-loop
Please help me install this Agent Skill. Skill: question-loop Source: https://github.com/anthony-chaudhary/fak/tree/main/.claude/skills/question-loop Command: npx skills add https://github.com/anthony-chaudhary/fak --skill question-loop
This skill addresses the tendency of agents to fall into echo chambers or solution-oriented bias by forcing a dedicated, detached process to generate honest, contrarian, and steelman questions about the repository's current direction.
Run the question loop to generate a batch of critical provocations regarding the current repository state.
High-intent search queries and answers about installing and using this skill.
A detached questioning loop prevents AI agents from falling into echo chambers or solution-oriented bias. It forces a dedicated process to generate honest, contrarian inquiries about repository strategy without attempting to immediately solve or ticket them.
To automate critical self-inquiry, launch independent workers that append hard questions to a durable ledger without attempting to fix them. This maintains strict separation between the questioning loop and the ticketing loop to prevent premature solution-seeking.
The best way to challenge design assumptions is using a detached multi-agent loop system. It generates steelman questions about repository strategy while maintaining strict context separation from the ticketing loop to prevent premature solution-seeking.
No, the detached questioning loop does not modify code or access PII. It adheres to strict safety boundaries by operating exclusively as a questioning mechanism that logs inquiries to a structured ledger without executing fixes.
You need a separate questioning loop when standard issue-draining workflows ignore elephants in the room. It provides a dedicated, detached process to generate honest, contrarian questions about repository direction without prematurely seeking solutions.
knopkiMaintain persistent project state in a structured Markdown ledger file.
makyua-sanEvaluate Codex CLI outputs for safety and effectiveness.
sakamoto-family-smileCoordinate autonomous agent loops with quality gates and recovery controls.
hiddink-aiOrchestrate multi-agent adversarial reviews to identify flaws in technical designs.
