qiaomu-opencli-oneshot

Pass

Audited by Gen Agent Trust Hub on Apr 9, 2026

Risk Level: SAFECREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill includes a hardcoded Bearer token in one of the templates. While this specific string is widely known as a public guest token for the Twitter/X API, hardcoding such values is a security anti-pattern and can lead to unintentional credential exposure if modified.
  • [DATA_EXFILTRATION]: The provided instructions and templates guide the agent to extract sensitive session data, such as cookies and CSRF tokens, from a user's browser environment to be used in external network requests. This behavior is necessary for the skill's function but represents a data exposure risk.
  • [COMMAND_EXECUTION]: The workflow involves generating local TypeScript files and executing them via shell commands using tools like npm and the opencli utility. This presents a risk if the generated code or arguments are influenced by malicious or unexpected external input.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes data from arbitrary, untrusted URLs provided at runtime.
  • Ingestion points: The process uses browser navigation and network request tools to retrieve content and API responses from external sites.
  • Boundary markers: No specific delimiters or instructions are provided to the agent to distinguish between valid data and potentially malicious instructions embedded in the captured web data.
  • Capability inventory: The skill utilizes browser evaluation, network fetching, and local file system writes, combined with shell command execution.
  • Sanitization: The skill does not implement sanitization or validation of the data captured from external websites before using it to generate adapter code.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 9, 2026, 06:58 AM