The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill configures and runs an external package from a public registry at runtime.
Evidence: The setup flow adds a command to .mcp.json that uses npx -y todoist-mcp-server to fetch and execute the server code.
[EXTERNAL_DOWNLOADS]: The skill depends on a remote package from the npm registry that is not from a verified trusted organization.
Evidence: Reference to todoist-mcp-server in the MCP configuration step.
[CREDENTIALS_UNSAFE]: The skill handles sensitive authentication tokens and persists them in multiple local configuration files.
Evidence: The skill prompts for a Todoist API token and writes it to both System/integrations/config.yaml and .mcp.json under the env section.
[COMMAND_EXECUTION]: The skill executes shell commands to validate connectivity to the external service.
Evidence: Execution of curl -s -H "Authorization: Bearer $API_KEY" https://api.todoist.com/api/v1/projects to test the token.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by ingesting untrusted task data from an external API.
Ingestion points: Task titles and descriptions fetched from Todoist via the todoist-mcp-server or curl calls.
Boundary markers: Uses a [dex:task-ID] marker for loop prevention, but lacks explicit delimiters or instructions for the model to ignore embedded commands in task data.
Capability inventory: The skill has the ability to execute shell commands (npx, curl) and modify local configuration files.
Sanitization: No evidence of sanitization or validation of the task content retrieved from the Todoist API before it is processed by the agent.

todoist-setup