The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to use npx @sendblue/cli, which downloads and executes the Sendblue CLI tool from the npm registry at runtime.
[EXTERNAL_DOWNLOADS]: Fetches and runs the @sendblue/cli package. As this is a resource provided by the vendor, it represents the intended functionality of the skill for sending notifications.
[INDIRECT_PROMPT_INJECTION]: The skill processes task outputs and duration to generate notification messages, creating a surface for potential data poisoning if malicious instructions are embedded in task results.
Ingestion points: Task execution status, command outputs, and environment variables like CLAUDE_TURN_DURATION_SECONDS (documented in SKILL.md).
Boundary markers: The skill provides templates for notification copy but does not define explicit delimiters for ingested data.
Capability inventory: Shell command execution via the npx @sendblue/cli send command.
Sanitization: The skill advises against including sensitive data in notifications but does not specify technical sanitization or escaping of the task output content.
[PERSISTENCE]: Provides instructions for registering Stop hooks in project-scoped configuration files (e.g., .claude/settings.json). While this is the primary purpose of the skill for automation, it creates persistent command execution triggered by agent lifecycle events.

sendblue-notify