Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This collection provides a set of eval-informed critical thinking frameworks designed for AI agents and automation builders. By implementing structured mental models, developers can guide tools like Claude Code or Cursor to handle complex technical tasks with greater precision.
The skills focus on grounding AI responses in evidence and establishing clear thresholds for investigation, helping agents recognize the boundaries of their own competence to avoid hallucinations.
WHO IT'S FOR
AI agent / automation builders
implementing critical thinking frameworks in AI agents
Technical leads / architects
reducing AI confabulation in complex technical tasks
Platform / DevEx teams
standardizing AI reasoning patterns across development tools
AI agent power users
guiding AI through first-principles problem solving
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Installs: Complete repository skills bundle copied into one project's .agents/skills directory.
Before you start
Git must be available.
Run from: A directory in which the cc-thinking-skills clone should be created.
Run this complete terminal command block from the directory where the clone should be created. Replace /path/to/project in both occurrences with the target project's path; leave the repository directory name unchanged.
Known technical, time, financial, reputational, and dependency costs of reversing it
Any deadline or opportunity that waiting could permanently close
Use the thinking-reversibility skill to assess this decision: [describe the decision]. Identify a concrete undo path, classify it as Type 2, Type 1.5, or Type 1, compare the downside and recovery cost of acting with the permanent loss from waiting, and suggest one option-preserving move. Finish with the appropriate process depth and a proposed commitment; do not execute the decision.
A description of the decision, incident, or subsystem
Observed cause-and-effect evidence
Urgency and whether small probes are safe
Use the thinking-cynefin skill to classify this problem: [describe the decision, incident, or subsystem]. Based on the available cause-and-effect evidence, predictability, urgency, and probe safety, assign one domain—clear, complicated, complex, chaotic, or disorder. Return the matching response mode, one to three first actions, and a falsifier that would force reclassification. If the problem is mixed, decompose it into separate units first.
Relevant evidence, incentives, trade-offs, and failure concerns
Use the thinking-steel-manning skill to evaluate this proposal: [proposal]. State the real alternative, extract the opposition's legitimate core concern, construct the strongest faithful opposing case using the supplied evidence, and name one concrete observation that would overturn the preferred position. Then engage that case and conclude with accept_opposing, revise, or reaffirm, including what changed and any residual risks.
Claude
Codex
Cursor
Gemini CLI
Copilot
DeepSeek