IKKF Manual
Overview
IKKF (I Know Kung Fu) is a runtime skill synthesis protocol for AI agents. It provides a structured approach to task execution through seven explicit phases.
Unlike other frameworks that hide their reasoning, IKKF makes every step visible and verifiable. Each phase produces concrete output that can be reviewed, audited, and improved.
Key Principles
- Explicit over implicit: Every step is stated, not assumed
- Verifiable output: Each phase produces visible results
- Memory-backed: Decisions and research are recorded
- Chunked work: Tasks broken into 10-15 minute pieces
Installation
macOS & Linux
Windows
Verify Installation
IKKF CLI v3.0.0
Commands
ikkf init
Initialize IKKF configuration. Creates the config file and task directory.
# Options:
# --force, -f Overwrite existing config
ikkf start
Start a new task with IKKF protocol. Launches the 7-phase process.
# Options:
# --confidence, -c Set confidence threshold (default: 50)
# --yes, -y Skip confirmation prompts
ikkf status
Check the status of the current task.
ikkf version
Display the current IKKF version.
The 7 Phases
Every IKKF task follows these seven phases in order. No skipping. No shortcuts.
1. Acknowledge
Confirm task understanding. State it back. Remove ambiguity.
2. Domain Detection
Identify the domain. Map required expertise. Load relevant skills.
3. Check Existing
Search memory. Find prior art. Check for existing solutions.
4. Research
Deep dive into best practices. 2026 standards. Trade-offs.
5. Decomposition
Break into 10-15 minute chunks. Clear deliverables. Map dependencies.
6. Synthesize
Plan implementation from research. Architecture decisions. Build order.
7. Execute & Persist
Build each chunk. Test. Commit. Record to memory. Deliver.
Configuration
IKKF stores its configuration in ~/.config/ikkf/config.yaml.
version: "3.0.0"
confidence_threshold: 50
chunk_size_minutes: 15
research_depth: normal
tasks_dir: ~/.config/ikkf/tasks
skills_dir: ~/.openclaw/workspace/skills/generated
Configuration Options
confidence_threshold— When to trigger deep research (default: 50)chunk_size_minutes— Target duration for task chunks (default: 15)research_depth— quick, normal, or thoroughtasks_dir— Where task files are storedskills_dir— Where generated skills are saved
Examples
Example 1: Building a Feature
# IKKF will guide you through:
# 1. Acknowledge: Confirm JWT auth requirements
# 2. Domain Detection: Backend/API development
# 3. Check Existing: Look for auth patterns in codebase
# 4. Research: JWT best practices, libraries
# 5. Decomposition: Break into chunks (setup, routes, middleware, tests)
# 6. Synthesize: Plan implementation order
# 7. Execute: Build, test, commit, document
Example 2: Research Task
Example 3: Debugging
Troubleshooting
Command not found
Ensure the IKKF binary is in your PATH. The install script adds this automatically, but you may need to restart your shell.
Config not initializing
Check that your home directory is writable and that the config directory doesn't already exist (unless using --force).
Task not starting
Ensure you've run ikkf init first. IKKF needs a configuration before it can track tasks.