The Agent Skills Directory

[SAFE]: The skill includes a shell script scripts/setup-windsurf.sh designed for developer onboarding. It performs benign environment checks, such as verifying the Windsurf version and installing standard development extensions (Prettier, ESLint) through the official CLI.
[SAFE]: The provided MCP configuration (mcp_config.json) utilizes npx to run official Model Context Protocol servers for PostgreSQL and GitHub. These are standard integrations for the Windsurf environment.
[SAFE]: Secret management is handled correctly using environment variable placeholders (e.g., ${GITHUB_TOKEN}, ${STAGING_DATABASE_URL}) rather than hardcoding credentials.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an environment where the agent reads instructions from .windsurfrules files committed to the repository. This is the primary intended functionality for Cascade AI's context management.
Ingestion points: Rules are ingested from .windsurfrules and .windsurf/rules/*.md files.
Boundary markers: The skill recommends clear headers and structured sections to define project stacks and architecture rules.
Capability inventory: The skill utilizes Read, Write, and Edit tools to manage these configuration files.
Sanitization: Instructions are treated as system-level guidance for the AI's behavior within the workspace.

windsurf-multi-env-setup