Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Ai Company Cfo

v2.0.0

AI公司首席财务官(CFO)技能包。财务规划、现金流管理、融资战略、资本配置、AI成本核算、动态预算、熔断机制。

0· 110·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
high confidence
!
Purpose & Capability
The skill is a CFO automation and legitimately needs file read/write and network/API access to talk to ERP, bank, cloud-billing and blockchain gateways — those permissions are present. However, the SKILL.md describes OAuth2/API-key authenticated endpoints, bank APIs, multisig/chain payments and cloud billing integrations while the registry metadata lists no required environment variables or credentials. That mismatch (no declared API keys/credentials for services the skill claims to integrate with) is unexpected and incoherent.
!
Instruction Scope
The instruction files direct the agent to read many workspace artifacts (SOUL.md, USER.md, memory/*) and to perform automated financial actions including chain payments, multisig execution, and automated fund allocation. AGENTS.md even suggests deleting BOOTSTRAP.md after following it. Combined read/write file permission + network API permissions mean these textual instructions would allow data access and potentially destructive actions if the agent is allowed to act — the instructions go beyond passive advising into operational execution without specifying how credentials/consent are managed.
Install Mechanism
This is an instruction-only skill with no install spec and no code files to run. That reduces the risk of arbitrary code being dropped during install. Static scanner had nothing to analyze.
!
Credentials
The skill claims integrations that normally require multiple sensitive credentials (bank API keys, cloud billing credentials, blockchain wallet/multisig keys, OAuth2 tokens), yet requires.env lists none. The frontmatter declares network: [api] and files: [read, write], which together permit reading local files (where users often store secrets) and sending them over the network. Requiring zero declared secrets for a capability that clearly needs them is disproportionate and raises the risk of ad-hoc credential usage or prompting the agent to ask for secrets at runtime.
!
Persistence & Privilege
always:false (good) but the skill requests mcp permissions: sessions_send and subagents. Those permissions allow the skill to create/send sessions to subagents (effectively spawn nested agents) which increases blast radius when combined with file read/write and network access. There is no justification in the metadata for creating subagents; this is a significant privilege and should be restricted unless absolutely needed.
What to consider before installing
This skill appears to be a coherent CFO advisor/automation, but there are important red flags you should address before installing: - Credentials: The skill describes connecting to ERP, bank APIs, cloud billing and blockchain wallets but declares no required credentials. Do not provide broad system credentials casually. Prefer scoped service accounts, read-only API keys where possible, and separate credentials per integration. - Network + file permissions: The skill is allowed to read/write workspace files and access APIs. Ensure no sensitive secrets (SSH keys, cloud creds, bank API keys, .env files) are stored in the workspace. Run in a sandboxed environment first. - Subagent/mcp privileges: The skill requests sessions_send and subagents. That allows it to spawn subagents which can expand capabilities and exfiltrate data. Disable subagent creation unless you explicitly trust the publisher and need that feature. - Deletion instruction: AGENTS.md suggests deleting BOOTSTRAP.md after following it. Any instruction that suggests deleting files should be treated cautiously — verify what would be deleted and why. - Source trust: The source is 'unknown' and owner metadata is minimal. If you plan to use this for real financial operations (payments, transfers), verify the publisher, review all endpoints the skill will call, and perform an operational security review of each integration (audit logs, least-privilege credentials, multisig manual approvals, and test in staging with no real money). Recommended actions: test in an isolated sandbox; require manual approval for any action that would move money; restrict or remove mcp subagent permissions; only provide minimal, scoped credentials; and review all runtime endpoints the agent will call before granting network access.

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

ai-companyvk97fnzn7fah8m80mxfy88ph2pd84n5fkc-suitevk97fnzn7fah8m80mxfy88ph2pd84n5fkcfovk97fnzn7fah8m80mxfy88ph2pd84n5fklatestvk9744eva18vdyj7zz6naanzt7x84vajt

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments