Skill Awrshift — AI skill for Claude Code
Adaptive decision framework for AI agents.
How to install Skill Awrshift
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open awrshift/skill-awrshift and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Skill Awrshift does
Adaptive decision framework for AI agents. One dynamic flow, user checkpoints at every phase, sandbox-safe experiments.
Alternatives in AI
- Flow Define — Multi-AI requirements scoping using Codex and Gemini CLIs (Double Diamond Define phase) 2.8k ★
- MioIsland — macOS Dynamic Island for AI coding agents 388 ★
- PaperFarm — Let AI agents run experiments in any repo while you sleep 353 ★
README
   
AWRSHIFT v2
**Adaptive decision framework for AI agents. User checkpoints at every phase.**
*Works with any AI coding assistant that supports [Agent Skills](https://agentskills.io)*
Why
AI agents jump to implementation. Most failures come from building the wrong thing. AWRSHIFT makes your agent research, define metrics, factcheck, and test in a sandbox — all before touching your main project. One dynamic flow adapts to any task complexity.
What's New in v2
| v1.0 | v2.0 |
|---|---|
| 3 modes (Quick/Standard/Scientific) | 1 dynamic flow — scope adapts per phase |
| Text-based questions | AskUserQuestion — structured A/B/C/D choices at every checkpoint |
| No metrics phase | EVALUATE-DESIGN — mandatory success criteria before planning |
| Factcheck only in Scientific | FACTCHECK — mandatory for all scopes |
| No sandbox rules | 10 safety rules — experiment sandbox is isolated from main project |
| No implementation gate | Double gate — DECIDE(GO) + file-by-file preview before touching main project |
How It Works
One flow. User controls depth.
IDENTIFY → RESEARCH → EVALUATE-DESIGN → HYPOTHESIZE → PLAN → FACTCHECK → TEST → DECIDE → [IMPLEMENT]At **every phase transition**, the agent:
- Tells you what was done
- Explains what happens next
- Asks you to choose (A/B/C/D or your own direction)
You're always in control. The agent never proceeds silently.
Related Skills
Safeshell
Safe shell operations with automatic checkpoints for AI agents. Let agents run freely. Everything is reversibl
Flow Deliver
Multi-AI validation, scoring, and review using Codex and Gemini CLIs (Double Diamond Deliver phase)
Flow Develop
Multi-AI implementation using Codex and Gemini CLIs (Double Diamond Develop phase)
Inclusive Design Skills
Inclusive design skills for AI coding agents — from cognitive accessibility to adaptive interfaces, inclusive
Jev Ultrafast MCP
Hand the browser work off: an MCP server where a decision model drives the page for your agent, so a flow cost
Drive Product Phase
Product Owner + Project Manager co-drive the team of role agents through the AI-PDLC lifecycle to a user-speci
Related Agents
Resource Optimizer
PROACTIVELY use this agent for dynamic resource monitoring and optimization. Specializes in performance mainte
Spec Kit Partner
A truly conversational, adaptive, and agentic subagent that partners with humans to co-create rigorous, multi-
Adaptive Mentor
Dynamic learning style optimization with real-time pedagogical adaptation. Provides personalized mentoring tha
