The Agent Skills Directory

[DATA_EXPOSURE]: The skill documents best practices for credential management by instructing the agent to read secrets from environment variables rather than hardcoding them (as seen in the .env section of SKILL.md).
[COMMAND_EXECUTION]: The skill utilizes standard curl and jq commands to interact with the Make.com REST API. These commands are localized to the documented well-known service infrastructure (make.com) and are standard for the skill's primary purpose of API operation.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data by retrieving scenario blueprints from the Make.com API. While this introduces a data ingestion surface, the risk is inherent to the skill's primary purpose of managing automation flows, and no malicious exploitation patterns were detected.
Ingestion points: REST API endpoints such as GET /scenarios/{id}/blueprint and GET /templates/{id}/blueprint (referenced in SKILL.md and references/blueprint-round-trip.md).
Boundary markers: None explicitly defined for API response processing.
Capability inventory: Includes capabilities for creating, updating, running, and deleting scenarios via curl network calls.
Sanitization: The skill promotes a blueprint 'round-trip' method where JSON payloads are manipulated and re-sent as strings, without explicit sanitization of embedded values.
[REMOTE_CODE_EXECUTION]: The skill references the integromat/make-mcp-server but does not include instructions for unverified downloads or automatic execution of remote scripts.

make