T09 · Insecure Skill Coding Practices
- Location
roundgame-agent.json:15- Finding
Bearer API Key May Be Redirected to an Untrusted Service Endpoint
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is coherent for a simulated-game API helper, but it gives an agent bearer-key access to state-changing betting endpoints with an unvalidated endpoint override and no explicit confirmation step.
Install only if you intend to let the agent use an LT99 API key for this simulated-game account. Before any authenticated call, verify the resolved service URL is the official LT99 endpoint, avoid setting generic SERVICES_BASE_URL in the agent environment, and require the agent to ask before /sim/enter or /sim/bet actions.
roundgame-agent.json:15Bearer API Key May Be Redirected to an Untrusted Service Endpoint
The skill depends on curl and instructs the agent to make HTTP requests to an externally configurable base URL, including authenticated requests carrying a Bearer API key. External transmission is expected for this skill's purpose, but it is still security-relevant because the base URL can be overridden by environment variables, increasing the chance of credential exfiltration or requests being redirected to an untrusted endpoint.
homepage: https://github.com/luyao-inc/LT99
requires:
anyBins:
- curl
---
# $LT99 模拟场(API Key)
The skill instructs the agent to send a user-provided Bearer API key to an external service and to place bets on the user's behalf, but it does not require an explicit warning, confirmation, or consent checkpoint before transmitting credentials or initiating wagering actions. This creates a real risk of unauthorized account use, accidental betting, and financial loss, especially because the documented execution flow normalizes direct bet placement as a routine agent action.
The skill purpose is broadly framed as using a bearer API key to perform simulated-game actions, but it does not define sufficient trigger constraints such as requiring explicit user confirmation per action, limiting when betting may be initiated, or narrowing permissible use cases. In an agent setting, this can enable the model to take sensitive account actions too readily once a key is present, increasing the risk of unauthorized or accidental transactions.
The manifest’s user-facing natural-language content is entirely in Chinese and provides no indication that language choice is optional or region-specific. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy violation.