The Agent Skills Directory

[SAFE]: The skill is an instructional manual (SKILL.md) for the OpenCLI project. It contains documentation and code templates rather than executable logic.
[COMMAND_EXECUTION]: Describes the intended use of CLI tools like opencli explore, opencli record, and opencli generate. These are part of the documented framework's core functionality for web automation.
[REMOTE_CODE_EXECUTION]: Provides templates for using page.evaluate to interact with web APIs. This is a standard and required practice for browser-based automation and data extraction.
[DATA_EXFILTRATION]: Outlines strategies for handling session cookies and authentication headers (e.g., Tier 2 and Tier 3 strategies) specifically for interacting with target websites as requested by the user. No unauthorized data transmission patterns were found.
[PROMPT_INJECTION]: Includes specific defensive coding guidance, instructing the agent to use JSON.stringify to escape variables when interpolating into browser evaluation strings to prevent code injection.
[DYNAMIC_EXECUTION]: Describes an architecture where TypeScript adapters are dynamically registered from a specific directory (clis/). This is presented as a standard extension mechanism for the OpenCLI tool.

qiaomu-opencli-explorer