The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill connects to and references official resources from https://www.superagent.sh. This includes the MCP server endpoint, documentation, and REST API for security scanning.
[COMMAND_EXECUTION]: Provides standard CLI commands for integrating the service with MCP clients, such as claude mcp add and modifications to mcp.json or config.toml.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by design, as it is intended to ingest and analyze untrusted external context including web pages, emails, and files.
Ingestion points: REST API endpoints for web pages, files, emails, and skills (/context/*).
Boundary markers: Explicitly defines a verdict policy (safe, caution, suspicious, dangerous) to handle scanner results and requires manual confirmation for high-risk actions.
Capability inventory: Includes shell command execution for lifecycle hooks and network operations to the Superagent API.
Sanitization: Instructs users to parse input as structured JSON and strictly forbids the use of eval or unsafe shell interpolation when processing tool input.
[DATA_EXFILTRATION]: Contains strong defensive instructions against credential exposure, explicitly warning against pasting API keys or webhook secrets into chat, logs, or source control, and recommending environment variables for storage.

superagent