The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing the lumail CLI tool via npx or bun. It also instructs the agent to use shell redirection to read token files (e.g., TOKEN=$(tr -d '\n' < ~/.config/lumail/token)).
[DYNAMIC_EXECUTION]: The skill includes a fallback for stale CLI installations that involves navigating to a hardcoded local path (/Users/melvynx/Developer/projects/lumail.io) and executing a script directly using bun src/cli/index.ts. This is environment-specific and executes local code.
[CREDENTIALS_UNSAFE]: The skill manages sensitive API tokens and OAuth credentials stored in ~/.config/lumail/. It provides explicit instructions for displaying raw tokens using the --raw flag (e.g., npx lumail auth show --raw and lumail accounts show <name> --raw), which exposes secrets in the agent's output.
[INDIRECT_PROMPT_INJECTION]: The skill includes a fetch_web_page tool and facilitates the creation of email content based on external data, creating a vector for instructions embedded in external content to influence the agent.
Ingestion points: The fetch_web_page tool (mentioned in the V2 Tools API section) and the processing of TipTap JSON documents that may contain remote content or URLs.
Boundary markers: None mentioned; instructions do not advise the agent to ignore embedded commands in fetched data.
Capability inventory: The skill has extensive write and network capabilities, including sending transactional emails (emails send), scheduling marketing campaigns (schedule_campaign), and configuring automation workflows (configure_workflow_v2_draft).
Sanitization: No sanitization or validation steps are described for data fetched from the web before it is used in email content or workflow configurations.

lumail