opencli-adapter-author
Pass
Audited by Gen Agent Trust Hub on May 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns detected. The skill's primary function is to guide the agent in authoring and testing JavaScript adapters for web scraping.
- [PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection as it processes data from external websites during API discovery. However, the workflow is strictly defined through decision trees and templates, minimizing the risk of the agent deviating from its task. Mandatory evidence chain:
- Ingestion points: "opencli browser network", "opencli browser analyze", and HTTP response bodies from target sites.
- Boundary markers: Absent, but the skill uses structured extraction and verification steps.
- Capability inventory: "Bash(opencli:*)" tools, file system access ("Read", "Write", "Edit"), and JavaScript execution within the OpenCLI environment.
- Sanitization: Documentation in "references/site-memory.md" explicitly instructs the agent to strip cookies, tokens, and personal identifiers (PII) before saving data fixtures.
- [DATA_EXFILTRATION]: The skill documentation specifically mandates the removal of sensitive information like cookies and private user data before persisting results to local storage. No suspicious network operations targeting non-whitelisted domains for data theft were found.
- [REMOTE_CODE_EXECUTION]: While the skill involves the generation and execution of JavaScript adapters, this is performed locally via the vendor-supplied "opencli" tool for the purpose of development and verification, matching the stated utility of the skill.
- [COMMAND_EXECUTION]: The skill leverages shell commands via "Bash(opencli:*)" for reconnaissance and debugging (e.g., "opencli doctor", "opencli browser network"). These operations are scoped to the development workflow of the OpenCLI ecosystem.
Audit Metadata