The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple Bash commands in its preamble for session management, update checks, and configuration loading from the user's home directory.
[REMOTE_CODE_EXECUTION]: Downloads and executes the Bun installation script from the official bun.sh domain if the runtime is missing. The process includes a checksum validation step to verify the integrity of the downloaded script.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources during the research phase, creating a potential injection surface when synthesizing design findings.
Ingestion points: Web search results and website snapshots/metadata via the browse tool (Phase 2).
Boundary markers: None explicitly implemented to isolate external content from instructions.
Capability inventory: Uses Bash, Write, Edit, and AskUserQuestion tools.
Sanitization: No explicit sanitization or filtering of external content is described before it is processed by the agent.
[DATA_EXFILTRATION]: Transmits telemetry data including skill usage, duration, and session identifiers to a remote endpoint. This behavior is subject to user opt-in via an interactive prompt during the first run.
[DYNAMIC_EXECUTION]: Uses eval and source on command output for environment setup and configuration. It also utilizes codex exec to dynamically generate design directions based on text prompts.

design-consultation