aoi-openclaw-security-toolkit-core
v0.1.6Fail-closed OpenClaw security toolkit (public-safe). Use to prevent accidental or unexpected data leakage by running local-only checks: default-deny allowlists, lightweight secret/token scans, egress-risk pattern scans, and prompt/document injection pattern scans. Use when preparing GitHub/ClawHub p...
Installation
AOI OpenClaw Security Toolkit (Core)
Why: Prevent “one bad commit” incidents (accidental file leakage + secret exposure) with a fast, local-only, fail-closed check.
When: Before committing/pushing, before publishing a skill, and when reviewing scripts/skills for unexpected egress behavior.
How: Run a single command to get PASS/WARN/BLOCK and an optional redaction-safe report.
Scope: Detection + reporting only (no auto-fix, no uploads, no auto-posting).
Quickstart: openclaw-sec check --preset repo --diff staged
This is a public-safe toolkit skill.
- Does: detect + report risks (PASS/WARN/BLOCK)
- Does NOT: auto-fix, auto-upload, auto-post, or exfiltrate data
CLI
Binary: openclaw-sec
Common:
openclaw-sec check --lang en
openclaw-sec check --lang ko
openclaw-sec scan-secrets
openclaw-sec scan-egress
openclaw-sec scan-prompt --file inbound.txt
Exit codes:
- 0 PASS
- 1 WARN
- 2 BLOCK
Default scan scope
If --paths is omitted, it scans existing paths among:
- .
- skills/
- scripts/
- context/
Rules
Rule files live in rules/:
- secret_patterns.txt
- egress_patterns.txt
- prompt_injection_patterns.txt
Edit these to tune sensitivity.