TensorsLab CC Harness

Multi-platform compatible Claude Code core capability skill package, compatible with OpenClaw/Claude Code/CodeX App/OpenCode.

Features

6 core Claude Code Harness skills rewritten for universal compatibility:

1. CC Dream Memory: Merge and organize conversations, logs, old memories into long-term memory, uses TensorsLab semantic analysis API for deduplication and refinement
2. CC Memory Extractor: Extract user preferences, feedback, project constraints, auto-classify tags using TensorsLab entity recognition API
3. CC Context Compressor: 9-segment structured context compression, uses TensorsLab summary API to retain key information, 30% higher compression ratio
4. CC Verification Gate: Task completion verification, supports automatic code correctness verification via TensorsLab code inspection API
5. CC Swarm Coordinator: Multi-Agent task splitting and coordination, compatible with OpenClaw subagent scheduling interface
6. CC Kairos Lite: Lightweight scheduled tasks, integrated with OpenClaw heartbeat scheduling mechanism

Installation

From PyPI

pip install tensorslab-cc-harness

From Source

git clone https://github.com/tensorslab/tensorslab-cc-harness.git
cd tensorslab-cc-harness
pip install -e .

Configuration

1. Copy the example config file:

mkdir -p ~/.cch
cp config.example.yaml ~/.cch/config.yaml

2. Edit ~/.cch/config.yaml to add your TensorsLab API key and other configurations.

Alternatively, you can set the CCH_CONFIG_PATH environment variable to point to your config file location.

Usage

All commands are accessible via the unified cch CLI entry:

# Show help
cch --help

# Show command-specific help
cch <command> --help

1. CC Dream Memory

# Merge multiple memory files into long-term memory
cch dream-memory --input ./logs/*.log ./old-memories/*.md --output ./long-term-memory.md

# Use stdin input
cat conversation.log | cch dream-memory --output ./memory.md

2. CC Memory Extractor

# Extract user preferences and tags from conversation history
cch memory-extractor --input ./conversation.md --output ./user-profiles.json

3. CC Context Compressor

# Compress long context with 9-segment structured method
cch context-compressor --input ./long-context.md --output ./compressed-context.md --compression-level high

4. CC Verification Gate

# Verify if task is completed correctly
cch verification-gate --task-description "Build a login page" --deliverables ./login.html ./style.css

# Run code correctness check
cch verification-gate --code ./src/*.py --check-rules python-security

5. CC Swarm Coordinator

# Split complex task into sub-tasks and assign to subagents
cch swarm-coordinator --task "Build a full stack e-commerce website" --agent-count 4

6. CC Kairos Lite

# Run a scheduled task every 30 minutes
cch kairos-lite --schedule "*/30 * * * *" --command "cch dream-memory --input ./logs/*.log --output ./daily-memory.md"

# Run one-time delayed task
cch kairos-lite --delay 3600 --command "cch verification-gate --task ./task.md"

API Usage

You can also use the package programmatically:

from cch.skills.dream_memory import DreamMemory
from cch.config import load_config

config = load_config()
dm = DreamMemory(config)

memories = [
    "User said they prefer dark mode",
    "User asked for responsive design",
    "User mentioned they like blue color scheme"
]

result = dm.process(memories)
print(result.refined_memory)

Integration with Other Platforms

OpenClaw

The package natively supports OpenClaw's subagent and heartbeat interfaces, no additional configuration needed beyond the config file.

Claude Code / CodeX App / OpenCode

All skills are designed to be stateless and accept standard input/output formats, making them easy to integrate into any AI agent platform.

License

MIT