π¦ ClawHub
clawder-agent
by @iyang1016
A production-grade AI coding agent that enforces JT directives by verifying code, running parallel sub-agents, logging mistakes, and ensuring edit integrity.
βοΈ Configuration
Add to your ZooClaw config (~/.openclaw/config.yaml):
clawder:
enabled: true
# Verification requirements (mandatory before "done")
verification:
required: true
typeCheck: true
lint: true
test: true
# Sub-agent settings
subAgents:
enabled: true
maxFilesPerAgent: 5-8
mode: worktree # worktree, fork, or remote
# Memory system
memory:
enabled: true
autoExtract: true
gotchasLogging: true
# JT Directives
directives:
seniorDevOverride: true
editIntegrity: true
contextDecayAwareness: true
autonomousBugFixing: true
mistakeLogging: true
π Tips & Best Practices
Verification failing?
# Check if verification tools exist
ls -la node_modules/.bin/tsc node_modules/.bin/eslintIf missing, configure project first
npx tsc --init
npx eslint --init
Sub-agents not spawning?
# Check config
grep "subAgents:" ~/.openclaw/config.yamlEnsure task touches >5 files
clawder --dry-run --prompt "Count files to modify"
Memory not working?
# Run memory scan
clawder --memory-scanCheck memory directory
ls -la ~/.openclaw/memory/
TERMINAL
clawhub install clawder-publish