technical-walkthrough-diagram
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHCOMMAND_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADS
Full Analysis
- COMMAND_EXECUTION (HIGH): The skill executes an unverified bundled JavaScript file
./scripts/fluxa-cli.bundle.jsfor core functionality. Bundled scripts are opaque and could contain malicious logic that is difficult to audit. - CREDENTIALS_UNSAFE (HIGH): The skill explicitly targets and manages sensitive credentials, including JWT tokens and agent IDs, stored in
~/.fluxa-ai-wallet-mcp/.agent-config.json. The instructions direct the agent to handle these credentials without user intervention. - EXTERNAL_DOWNLOADS (MEDIUM): The skill communicates with untrusted domains
proxy-monetize.fluxapay.xyzandava-agent.fluxapay-qa.xyz. These are not included in the trusted external sources and represent a risk for data exfiltration or unverified tool responses. - PROMPT_INJECTION (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8) because it uses the
webfetch_magictool to ingest arbitrary content from external URLs without sanitization or boundary markers. This untrusted data is then used to generate diagrams and posts, which could allow an attacker to influence the agent's behavior via embedded instructions in fetched documentation. Mandatory Evidence Chain for Category 8: 1. Ingestion points:webfetch_magictool (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Filesystem access viafluxa-cli.bundle.js, shell execution, and network operations. 4. Sanitization: None identified.
Recommendations
- AI detected serious security threats
Audit Metadata