The Agent Skills Directory

[SAFE]: The skill utilizes vendor-specific libraries (@agentrhq/webcmd/registry and @agentrhq/webcmd/errors) for its core functionality, which is standard for specialized agent tools.
[INDIRECT_PROMPT_INJECTION]: The skill involves processing untrusted data from external websites via browser automation. This creates a surface for indirect prompt injection where instructions could be hidden in web content. However, the skill provides structured workflows for field decoding and verification that focus on data extraction rather than instructional execution, and it explicitly instructs agents not to bypass safety or risk controls.
Ingestion points: Browser-driven web content accessed via webcmd browser * commands (references/api-discovery.md).
Boundary markers: The instructions guide the agent to use structured field mapping and webcmd browser verify fixtures to validate data, which acts as a verification layer.
Capability inventory: The skill has capabilities to Write to the local filesystem (~/.webcmd/sites/) and execute Bash commands scoped to the webcmd toolset.
Sanitization: The skill emphasizes structural verification and comparison against visible page data to ensure accuracy and reduce the risk of processing unintended content.
[COMMAND_EXECUTION]: The skill facilitates the creation of new Javascript-based adapters in the user's local directory (~/.webcmd/clis/). This is the primary intended purpose of the tool and is performed through established registry patterns.

webcmd-adapter-author