The Agent Skills Directory

[SAFE]: The skill documentation follows standard development and testing practices for automating Electron applications using the --remote-debugging-port flag.
[DATA_EXFILTRATION]: The skill facilitates reading and extracting data from application user interfaces (e.g., extracting JSON state or taking screenshots). This is the primary intended purpose of the automation tool and does not involve unauthorized exfiltration to external domains.
[PROMPT_INJECTION]: By interacting with third-party applications like Slack, Discord, and VS Code, the agent is exposed to untrusted external content which could facilitate indirect prompt injection. This is a known surface for UI automation tools and is noted here as a characteristic of the tool's usage.
Ingestion points: UI data captured via agent-browser snapshot and agent-browser get text in the SKILL.md file.
Boundary markers: None specified in the instructions.
Capability inventory: The agent-browser tool allows clicking, filling forms, and navigating within the targeted applications.
Sanitization: None identified; the agent processes the raw UI output from the application.

electron