The Agent Skills Directory

[DATA_EXFILTRATION]: The skill implements strict data handling policies, explicitly forbidding the logging of credentials, tokens, cookies, or private settings. It provides specific redaction guidelines and encourages the use of sanitized snapshots when logging runtime objects.
[EXTERNAL_DOWNLOADS]: The skill mentions external tools such as playwright-cli and dependencies via npm or meteor add. It requires the agent to check for tool availability and obtain explicit user authorization before performing any installations or browser downloads.
[COMMAND_EXECUTION]: The skill uses standard development commands like meteor run --inspect and playwright-cli. Safety protocols are included, such as warnings against binding inspectors to untrusted networks and instructions to avoid mutating production data without authority.
[INDIRECT_PROMPT_INJECTION]: The skill handles user-provided URLs and application paths. Risks are mitigated by instructions that limit discovery to the provided scope and explicitly forbid scanning unrelated directories, processes, or ports.
Ingestion points: User-provided URLs and application paths.
Boundary markers: Scope limitation to workspace and explicit request for URLs.
Capability inventory: Shell command execution via meteor and playwright-cli.
Sanitization: Strict log redaction rules and focus on observable UI results.

meteor-debugging