The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes bundled Python scripts, validate-findings.py and serve-report.py, using uv run to perform validation of the findings JSON and to host the local report viewer.
[EXTERNAL_DOWNLOADS]: Navigates to and processes external, user-provided URLs via the agent-browser tool to execute QA checks and capture evidence.
[DATA_EXFILTRATION]: Accesses and processes browser console logs and network request data. The skill contains specific instructions for the agent to filter out sensitive information like cookies, tokens, or private data before including them in the final report.
[SAFE]: Starts a local report server bound to 127.0.0.1:10891. The server implementation in scripts/serve-report.py includes robust path validation and sanitization to prevent directory traversal and ensure it only serves files from the designated temporary report directory.
[SAFE]: Utilizes a structured evidence contract (JSON) and an HTML report viewer that implements HTML entity escaping for all dynamic content, reducing risks associated with indirect prompt injection from analyzed web pages.

browser-qa