π¦ ClawHub
Feedback-Loop-v2
by @danielfoojunwei
A self-improving feedback loop skill that works fully standalone OR integrates with intent-engineering and dark-factory when available. Observes any system o...
π‘ Examples
Standalone β any JSON log
python scripts/orchestrator.py --input my_execution_log.json --goal "Process tickets in under 2 minutes"
Standalone β plain text description
python scripts/orchestrator.py --text "Script ran but missed 3 null input cases and took 4 minutes" --goal "Handle all inputs"
Standalone β continuing a prior cycle
python scripts/orchestrator.py --observation observation.json
Dark Factory Enhanced
python scripts/orchestrator.py --outcome outcome_report.json --goal "Achieve 98% test pass rate"
Full Triad (after intent-engineering + dark-factory)
python scripts/orchestrator.py --outcome outcome_report.json --spec specification.json
Run stages independently
python scripts/observer.py --input log.json --goal "..." --output observation.json
python scripts/analyzer.py --observation observation.json --output analysis.json
python scripts/orchestrator.py --analysis analysis.json --output-dir ./reports/
βοΈ Configuration
All configuration lives inside the skill β no external files required.
references/alignment_values.json β Edit to define your organization's goals and values. The analyzer checks every observation against these values to produce the alignment score.
references/scoring_weights.json β Edit to change how the performance score is calculated (e.g. weight pass rate more heavily than speed).
references/suggestion_rules.json β Edit to add custom rules for generating improvement suggestions.
TERMINAL
clawhub install auto-feedback