A premium, production-quality collection of structured AI engineering skills designed to automate and orchestrate the entire Software Development Life Cycle (SDLC) for AI Coding Agents.
The AI Skill Framework provides a standardized, modular set of instruction sheets (Skills) that AI Coding Agents can load to perform architectural, planning, engineering, and release tasks. It utilizes a feature-centric workflow where every feature is assigned a unique Feature ID tracked consistently from discovery to deployment.
To eliminate duplicated constraints and enforce consistent agent behavior across all tasks, the framework utilizes a centralized policy file: AI_RULES.md.
This file defines the global rules for:
- Approval Gate Policy: Ensures no state changes happen without user verification.
- Git Workflow Policy: Controls branch naming, switches, commits, pushes, and tags.
- Memory First Policy: Prioritizes local Project Memory over generic repository scans.
- RAG Policy: Sets strict priorities for indexing and retrieval sequences.
- Artifact Policy: Governs folder mapping and file prefix layouts.
- Versioning Policy: Enforces SemVer and tagging layouts.
- Documentation Policy: Standardizes formatting, templates, and relative path linking.
- Testing Policy: Regulates verification gates and compile/test verification.
- Release Policy: Specifies the sequential release execution loop.
Skills do not duplicate these common rules. Instead, they reference them at the top of their definition:
## 🔒 GLOBAL POLICY REFERENCES
This Skill MUST strictly follow the global policies defined in [AI_RULES.md](../../AI_RULES.md):
- **Approval Gate Policy** (Section 1)
- **Git Workflow Policy** (Section 2)- Creating/Modifying Policies: Edit
AI_RULES.mddirectly. Changes automatically govern all referencing skills. - Creating New Skills:
- Create a folder under
skills/<your-new-skill>/. - Write a
SKILL.mdusing the metadata frontmatter. - Reference the required global policies from
AI_RULES.mdin your skill's boundary section. - Register the new skill name in
MANIFEST.jsonand catalog it inSKILLS.md.
- Create a folder under
To monitor your active session progress, checkpoints, and token usage in real-time, you can install the companion AI Workflow Visualizer extension:
The framework includes a global CLI wrapper (aiwf) that allows you to trigger framework actions from any terminal and from any project directory without needing to specify the framework's clone path.
Clone the repository to a folder of your choice, navigate to it, and execute the bootstrap script for your platform:
On Linux/macOS:
./bootstrap.shOn Windows (PowerShell - run as Administrator if needed to update PATH):
.\bootstrap.ps1On Windows (CMD Command Prompt):
bootstrap.batThis will add the CLI path to your environment PATH variables.
Once bootstrapped, open a new terminal in any Git project and execute:
-
Install Framework Skills into Project:
aiwf install
Deploys
AI_RULES.md,MANIFEST.json,skills/, andtemplates/into the project's local.agents/folder. -
Update/Sync Framework Skills:
aiwf update
Synchronizes local skills with latest framework version without overwriting user customization.
-
Run Health Diagnostics:
aiwf doctor
Verifies framework installation, global wrappers availability, local folder integrity, and version metadata.
-
Check Framework Version:
aiwf version
Reports framework, CLI wrapper, and repository location/version.
-
Manage Project Memory:
aiwf memory <bootstrap | update | search> [options/query]
Handles full project memory bootstrapping, incremental updates via git-diff/timestamp, and semantic/keyword query searching.
-
Safely Remove Skills:
aiwf uninstall
Cleans up local framework skills and rules, keeping user memory files intact.
The framework operates on a linear, gated workflow to ensure software quality at every milestone (refer to the infographic at the top of the page).
ai-skill-framework/
├── docs/ # Standardized project documentation directories
│ ├── brainstorming/ # Requirement discovery and planning prompts (FEAT-XXX_*.md)
│ ├── plans/ # Approved project plans (FEAT-XXX_*_plan.md)
│ ├── designs/ # Technical design blueprints (FEAT-XXX_*_blueprint.md)
│ ├── issues/ # Bug fix specifications (FIX-XXX_*.md)
│ ├── quick/ # Quick feature specifications (QUICK-XXX_*.md)
│ ├── adr/ # Optional Architecture Decision Records (ADR-XXX_*.md)
│ ├── debug/ # Debug and Build Diagnostics (FEAT-XXX_debug.md)
│ └── verification/ # Final Quality Gate Reports (FEAT-XXX_verify.md)
├── skills/ # The core skill library
│ ├── blueprint-to-implementation/
│ ├── brainstorming/
│ ├── brainstorming-to-plan/
│ ├── create-adr/
│ ├── debug-to-verify/
│ ├── environment-bootstrap/
│ ├── environment-health/
│ ├── fast-fix/
│ ├── frontend-design/
│ ├── frontend-visual-debug/
│ ├── idea-to-planning-prompt/
│ ├── implementation-to-debug/
│ ├── implementation-to-release/
│ ├── initialize-workflow/
│ ├── plan-to-blueprint/
│ ├── planning-prompt-to-plan/
│ ├── project-discovery/
│ ├── project-memory-bootstrap/
│ ├── project-memory-update/
│ ├── project-rag-search/
│ ├── quick-feature/
│ ├── quick-fix/
│ ├── resume-workflow/
│ ├── software-development-workflow/
│ └── workflow-runtime/
├── bootstrap.sh # Global CLI bootstrapper for Unix
├── bootstrap.ps1 # Global CLI bootstrapper for PowerShell
├── bootstrap.bat # Global CLI bootstrapper for CMD
├── install.sh / install.ps1 # Local project installer
├── update.sh / update.ps1 # Local project updater
├── uninstall.sh / uninstall.ps1 # Local project uninstaller
├── doctor.sh / doctor.ps1 # Workspace diagnostics
├── version.sh / version.ps1 # Version inspector
├── CHANGELOG.md # Main release logs
├── LICENSE # MIT Open Source License
├── MANIFEST.json # Machine-readable skill registry metadata
└── SKILLS.md # Master index of all skills
- Name: Kyle Dang
- Email: kyleit@klexpress.net
- Website: https://www.klexpress.net
