Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
CFGPU API Skill
v1.1.0A powerful OpenClaw skill for managing and automating GPU container instances on CFGPU cloud platform. Designed for AI/ML developers, researchers, and conten...
⭐ 0· 54·0 current·0 all-time
byAIAD@r600a-code
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name, README, SKILL.md and scripts consistently implement a CLI for CFGPU API operations (region/gpu listing, instance lifecycle, image management). Requesting an API token is appropriate for the stated purpose.
Instruction Scope
Runtime instructions and shipped scripts perform expected API calls, but the helper builds curl commands as strings and calls eval in api_request(), which can enable command injection if inputs (instance names, image IDs, etc.) contain malicious characters. setup-env.sh reads and writes ~/.cfgpu/token and may append export lines to shell rc files — these actions modify user configuration and should be made explicit to users.
Install Mechanism
No install spec; skill is instruction-plus-shell-scripts only. No network downloads or archive extraction at install time are requested, which lowers installation risk.
Credentials
Only CFGPU_API_TOKEN (and optional CFGPU_API_TOKEN_FILE) are used — this is proportionate to a cloud API client. However the scripts automatically persist the token to ~/.cfgpu/token and may add export lines to ~/.bashrc or ~/.zshrc during setup-env.sh, which escalates the scope of environment changes and storage of credentials on disk.
Persistence & Privilege
always:false (normal), but setup-env.sh and package scripts will write files under the user's home (~/.cfgpu/token) and may append exports to shell rc files — persistent modifications to user config. package-for-github.sh references absolute /root/.openclaw/... paths and writes temporary packaging artifacts under /tmp; these assume filesystem access and could reveal host-specific paths.
Scan Findings in Context
[unicode-control-chars] unexpected: SKILL.md contained unicode control characters flagged as prompt-injection patterns. These may be benign formatting artifacts but could also be used to hide or obfuscate instructions; review SKILL.md for invisible characters before trusting contents.
What to consider before installing
This skill's functionality matches its description, but review before installing: 1) Inspect scripts (cfgpu-helper.sh, setup-env.sh, package-for-github.sh) yourself — they will read/write ~/.cfgpu/token and setup-env.sh can append an export to your shell rc. 2) The helper builds curl commands as strings and uses eval — avoid passing untrusted values (e.g., copy-paste instance names) and consider sanitizing inputs; this is a command-injection risk. 3) SKILL.md had unicode-control characters — open it in a hex-aware editor to ensure nothing hidden. 4) Run the skill in an isolated environment (container or dedicated account) first; create a token with minimal privileges and rotate it after testing. 5) If you don't trust the unknown publisher or cannot audit the scripts, do not install on a production workstation. If you proceed, consider modifying api_request to call curl directly (no eval) and avoid auto-appending credentials to shell rc files.Like a lobster shell, security has layers — review code before you run it.
latestvk972hgt50fr7hvdf8mgwbtbws983p10g
License
MIT-0
Free to use, modify, and redistribute. No attribution required.