Predict
PassAudited by ClawScan on May 1, 2026.
Overview
The skill appears purpose-aligned and local-only, with no credential or network use, but users should notice that it stores forecast records persistently and references helper files that are not included.
This skill looks safe to install from the provided artifacts if you are comfortable with local persistent forecast records. Before relying on it, note that several documented scripts are missing, and delete memory/predict files when you do not want forecast or risk-assessment history retained.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Forecast topics and periods may remain on the local machine after use.
The script persists forecast metadata locally in an OpenClaw memory directory. This is disclosed and purpose-aligned, but prediction and risk data may be sensitive and retained across sessions.
PREDICT_DIR = os.path.expanduser("~/.openclaw/workspace/memory/predict") ... json.dump(data, f, indent=2)Review and delete files under memory/predict when forecasts or risk assessments should no longer be retained.
Some advertised workflows may fail or rely on the model's own reasoning rather than reviewed helper code.
The SKILL.md documents several helper scripts, but the provided manifest includes only SKILL.md and scripts/forecast_trend.py. This is a functionality/provenance gap rather than evidence of malicious behavior.
`generate_scenarios.py` | `assess_risk.py` | `evaluate_probability.py` | `track_accuracy.py` | `build_model.py`
Treat the included forecast_trend.py as the only reviewed script and verify any missing helper files before relying on those workflows.