The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions to relaunch desktop applications with the --remote-debugging-port flag. This action opens a local network port that grants programmatic control over the application's interface and internal data through the Chrome DevTools Protocol (CDP).
[DATA_EXFILTRATION]: Facilitates access to and extraction of data from sensitive applications. The skill explicitly identifies compatibility with communication platforms (Slack, Discord, Signal) and password managers (1Password), creating a risk path for exposing private communications or stored credentials if the agent is misused.
[PROMPT_INJECTION]: The skill establishes a surface for indirect prompt injection (Category 8) by reading and interacting with untrusted data from communication apps.
Ingestion points: Untrusted data enters the agent context via agent-browser snapshot and agent-browser get text (SKILL.md).
Boundary markers: The skill does not implement delimiters or instructions to ignore embedded commands within the processed application data.
Capability inventory: The agent can click, type, and fill forms within applications using the agent-browser tool (SKILL.md).
Sanitization: No sanitization or validation of the content read from desktop applications is performed before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The configuration allows the use of npx agent-browser, which dynamically downloads and executes the agent-browser tool from the NPM registry at runtime.

electron