Security audit
Security checks for vulnerabilities and agentic risk
The skill is coherent for a roast API, but it also guides agents through wallet-based purchases and public posting, which should be reviewed before use.
Install only if you are comfortable with the agent calling an external roast API and, for paid use, working around a wallet-controlled on-chain purchase flow. Keep wallet tools disabled unless you explicitly approve the exact approve, purchase, signing, and claim action in-session, and require review before any generated roast is posted publicly.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
- PM2 for process management
- OpenAI API provides roasts
-
Secrets must be stored in .env (never committed).
------------------------------------------------------------
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
- PM2 for process management
- OpenAI API provides roasts
-
Secrets must be stored in .env (never committed).
------------------------------------------------------------
The file says agents must never use wallet functions without human confirmation, but immediately provides instructions for agents to purchase directly using a controlled wallet. This contradiction is dangerous because agent frameworks may implement the actionable purchase flow while treating the warning as advisory, enabling unauthorized or insufficiently consented on-chain spending.
The README presents conflicting guidance: it says agents must never use wallet functions without human confirmation, but elsewhere promotes direct wallet-controlled purchases by agents. This ambiguity is dangerous because agent integrators may implement autonomous financial actions without an explicit confirmation checkpoint, leading to unauthorized or unintended blockchain transactions.
The section describing agent/bot purchasing workflows encourages building approval and purchase transactions and purchasing directly with a controlled wallet, but does not repeat a clear safety warning at the point where the action is described. In agent ecosystems, operational instructions near actionable endpoints are often followed literally, so missing an in-context warning increases the chance of unauthorized spending or unsafe wallet automation.
The skill’s stated purpose is roast generation, but it also documents plan purchasing, transaction building, signature-based auth, and API-key issuance. Expanding a content-generation skill to include wallet and entitlement flows increases the attack surface and creates a path for unintended financial actions if an integrating agent follows the instructions too broadly.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
GET /health
curl:
curl -sS https://theroaster.app/health
2) Contract Metadata
GET /api/v1/contract
The skill instructs bots to post returned roast text unchanged as a public comment, but does not warn that this is third-party generated content being published externally. That creates a risk of reputational harm, harassment, policy violations, or accidental publication of unsafe output because the agent is told not to review, transform, or gate the content before posting.
No suspicious patterns detected.