ClawHub

ZT4AI Self-Audit

v1.0.0

Zero Trust security audit for AI agent workspaces, skills, and configurations. Based on Microsoft's Zero Trust for AI (ZT4AI) framework and the "Caging the A...

⭐ 0· 81·0 current·0 all-time
by@tanarchytan
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description match the delivered artifacts: audit checklists, classification guides, a report template, and a local integrity-check script. The directories and operations targeted (workspace and skill locations, checksums, grep for secrets, firewall/egress checks) are expected for an auditing tool.
ℹ
Instruction Scope
Instructions perform local discovery (ls, find, grep, sha256sum) and an outbound reachability test (curl https://httpbin.org/get). Those actions are consistent with auditing, but the network test will generate an outbound request — expected for an egress check but worth noting because it touches the network.
✓
Install Mechanism
No install spec and no remote downloads. The included shell script is local and self-contained (uses standard UNIX tools). This is the lowest-risk install model.
✓
Credentials
The skill requires no credentials or special environment variables. The script optionally reads OPENCLAW_WORKSPACE and HOME, which is appropriate for a workspace-targeted audit. No secret exfiltration or unrelated credential access is requested.
✓
Persistence & Privilege
always is false and the skill does not request elevated privileges or attempt to modify other skills or system-wide configs. It writes baselines/reports into the agent workspace (expected behavior).
Assessment
This skill appears coherent for auditing agent skills and workspace integrity. Before running: (1) inspect the included scripts (scripts/integrity-check.sh) to ensure they haven't been tampered with; (2) generate the initial baseline only after manual review of files you trust; (3) run the audit in a sandbox or with least-privilege user if possible (the curl test will make an outbound request to httpbin.org); and (4) treat the integrity baseline as sensitive — an attacker who can modify it can hide tampering. If you want higher assurance, run the checks on an isolated host or review the script line-by-line prior to execution.

Like a lobster shell, security has layers — review code before you run it.

latestvk97dc9evhctv6a7s01cg65fnx983a9v5

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments