The Agent Skills Directory

[PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by interacting with external Electron applications (e.g., Slack, Discord, VS Code). These apps display content from untrusted third parties which could include hidden instructions designed to manipulate the agent's behavior.
Ingestion points: Data is ingested via agent-browser snapshot and agent-browser get text as described in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing data from these apps.
Capability inventory: The skill has access to UI interaction tools (click, fill, press) and the Bash tool.
Sanitization: No evidence of sanitization or filtering of application content is present.
[COMMAND_EXECUTION]: The skill uses bash commands to launch desktop applications with debugging arguments (e.g., open -a 'Slack' --args --remote-debugging-port=9222) and to execute the agent-browser utility.
[EXTERNAL_DOWNLOADS]: The skill leverages npx to execute the agent-browser package, which involves fetching the tool from the npm registry.

electron