Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Ainative Agent Framework
v1.0.0Build multi-agent systems and swarms on AINative. Use when (1) Orchestrating multiple specialized AI agents, (2) Dispatching tasks to OpenClaw agents (aurora...
⭐ 0· 104·1 current·1 all-time
byToby Morning@urbantech
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The SKILL.md demonstrates networked use (X-API-Key headers, ACP gateway token) and CLI usage (openclaw, zerodb-rlhf-feedback, scripts/cody_openclaw.py) that are necessary for orchestrating OpenClaw/AINative agents, but the registry metadata declares no required env vars, no primary credential, and no required binaries. That mismatch (a network API and gateway token are clearly required by the instructions but not declared) is disproportionate and unexplained. The skill also lists internal config paths (.openclaw/openclaw.json, scripts/, src/) that are not included in the package.
Instruction Scope
Runtime instructions tell the agent to send API requests including X-API-Key and ACP tokens, run local binaries (openclaw), and execute subprocess calls to dispatch tasks. The instructions encourage use of sensitive tokens (YOUR_GATEWAY_TOKEN, ak_your_key) but do not define where those credentials come from or how they should be provided. The SKILL.md references local scripts and config files that are not present in the skill bundle, which could confuse users or lead to copy-pasting secrets into commands without context.
Install Mechanism
This is instruction-only (no install spec and no code files), which reduces the risk of arbitrary code being installed by the registry. However, the instructions assume external tooling (openclaw CLI, zerodb-rlhf-feedback, Python scripts) and an external API host (api.ainative.studio). Because nothing is installed by the skill itself, the security posture depends entirely on those external binaries and endpoints, which are not validated or linked from a known homepage.
Credentials
The SKILL.md uses API keys and gateway tokens in examples and headers (X-API-Key, YOUR_GATEWAY_TOKEN) but requires.env is empty and no primary credential is declared. That omission is a proportionality problem: a multi-agent orchestration skill legitimately needs at least an API key and possibly a gateway token, but the registry metadata does not request them. The skill also references other tooling (zerodb CLI) and local config paths without declaring access requirements.
Persistence & Privilege
always is false, the skill is user-invocable and allows model invocation (normal). The SKILL.md does not instruct the agent to modify other skills' configs or request permanent presence. It does describe using a memory API to persist memories, which is normal for agent frameworks but is conducted via external API calls rather than modifying agent platform config.
What to consider before installing
This skill appears to be an orchestration guide for AINative/OpenClaw agents, but the package omits key provenance and credential declarations. Before installing or using it: (1) Ask the author for source code or a homepage so you can verify the scripts and CLI it references (scripts/cody_openclaw.py, .openclaw/openclaw.json). (2) Expect to need an API key (X-API-Key) and a gateway token for ACP — do not paste production keys into demos; use scoped, revocable test keys. (3) Verify the external endpoints (https://api.ainative.studio) independently and prefer to run these tools in a sandboxed environment. (4) Confirm you have the referenced binaries (openclaw, zerodb-rlhf-feedback) from trusted releases. (5) If you cannot verify source or tooling, treat the skill as untrusted and avoid supplying real credentials or running commands that include tokens. Providing these additional details (declared required env vars, links to binaries/releases, or bundling the referenced scripts) would raise confidence.Like a lobster shell, security has layers — review code before you run it.
latestvk977mq9z9aa3c4sd8c89bhdc3183gjwd
License
MIT-0
Free to use, modify, and redistribute. No attribution required.