Back to skill
Skillv1.0.0
ClawScan security
Self-Upgrade (Tianyi) · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignFeb 28, 2026, 10:51 PM
- Verdict
- benign
- Confidence
- medium
- Model
- gpt-5-mini
- Summary
- The skill's requested actions (reading/modifying OpenClaw config, creating backups, running built-in upgrade/diagnostic commands) align with its self-upgrade purpose, but it manipulates local credentials in config and contains a hard-coded workspace path that merit review before use.
- Guidance
- This skill is coherent with a self-upgrade tool, but review before installing: (1) Inspect the full script to confirm it does not transmit tokens off-machine and logs sensitive values. (2) Confirm that migrating authToken fields is intended — back up current configs and verify rollback works. (3) Remove or parameterize the hard-coded skillsPath (D:\workspace\openclaw_ceo\skills) to avoid touching unrelated directories. (4) Understand what openclaw doctor --fix does on your system (run in DryRun first). (5) Prefer to require explicit user confirmation for medium/high-risk steps and restrict autonomous invocation for those operations. If you cannot review the script or do not trust it to handle secrets safely, do not enable automatic mode.
Review Dimensions
- Purpose & Capability
- okName/description (self-upgrade, config migration, backups, diagnostics) match the instructions: reading ~/.openclaw/openclaw.json, migrating fields, creating backups, running openclaw doctor and npm checks are expected for an upgrade helper.
- Instruction Scope
- noteInstructions explicitly read and overwrite the OpenClaw config, migrate authToken → auth.token, create backups, and run openclaw doctor --fix. These are within upgrade scope, but any operation that touches authentication fields warrants careful review and explicit authorization; the skill does not show safeguards for avoiding accidental exposure of secrets beyond prohibiting unauthorized API key changes.
- Install Mechanism
- okInstruction-only skill with no install spec or external downloads — lowest install risk. No code files to execute were bundled by the registry.
- Credentials
- noteThe skill requests no environment variables or external credentials, which is appropriate. However it reads and writes the local OpenClaw config that may contain API tokens/credentials; this is proportionate for a migration tool but increases sensitivity and should be justified/authorized by the user.
- Persistence & Privilege
- okalways is false and the skill is user-invocable. The ability to modify local configuration is expected for an upgrade tool; autonomous invocation is permitted by platform defaults — combine with user confirmation for non-low-risk steps as the SKILL.md recommends.