The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages a sensitive authentication token (LUCKIN_MCP_TOKEN). It provides instructions for the agent to store this token in a local plain text file at ~/.my-coffee/LUCKIN_MCP_TOKEN. Although the skill advises setting file permissions, storing credentials in plain text on disk increases exposure risk. Additionally, the bundled script scripts/check-env.sh reads platform-specific configuration files (~/.pi/agent/mcp.json and ~/.pi/agent/settings.json) which often contain sensitive authentication details for other services.
[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands, including a local diagnostic script (scripts/check-env.sh) and various curl commands to interact with external APIs.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to https://ipinfo.io/json for geolocation and https://gwmcp.lkcoffee.com/order/user/mcp for order management. These are documented functional requirements but represent external network dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external API responses without explicit isolation, creating a potential surface for indirect prompt injection.
Ingestion points: API responses from gwmcp.lkcoffee.com and ipinfo.io processed in SKILL.md.
Boundary markers: None explicitly defined to isolate external data from the skill's own instructions.
Capability inventory: File system access (read/write) for token management and shell command execution (curl, bash) for API interactions.
Sanitization: No documented sanitization or schema validation of API responses is performed before the data is used to influence the agent's behavior.

my-coffee