The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands for environment discovery and skill listing. Evidence: It executes grep on template.yaml, basename on the current path, and ls on .claude/skills/ to identify the agent and available procedures.
[SAFE]: The skill maintains a local registry for state persistence and interacts with the Trinity MCP server. Evidence: It reads and writes to ~/.schedule-registry.json using the included scripts/registry.py helper, and utilizes official Trinity MCP tools for schedule management.
[PROMPT_INJECTION]: An attack surface for indirect prompt injection exists due to the processing of external schedule names and local skill metadata. Ingestion points: Remote schedule data from Trinity MCP and local skill file metadata. Boundary markers: Absent in status reports. Capability inventory: Python script execution, local file modification, and remote tool invocation. Sanitization: Descriptive fields are displayed without content validation.

trinity-schedules