The Agent Skills Directory

[PERSISTENCE]: The skill is designed to establish long-term persistence by instructing the agent to create recurring scheduled tasks (crons) and maintain a persistent knowledge graph. It leverages the @anthropic/scheduled-tasks-mcp-server to execute prompts at regular intervals (e.g., daily standups, hourly monitoring) without direct user intervention.
[PRIVILEGE_ESCALATION]: The skill enables "Computer Use" capabilities via the @anthropic/computer-use-mcp-server. This facilitates high-privilege actions including browser automation, desktop control (mouse/keyboard interaction), and file system operations beyond the standard CLI scope.
[EXTERNAL_DOWNLOADS]: The setup instructions direct the installation of several MCP servers from a trusted organization using npx -y. This includes the memory, scheduled-tasks, and computer-use servers hosted under the @anthropic registry.
[INDIRECT_PROMPT_INJECTION]: The autonomous workflows are designed to ingest data from external, potentially untrusted sources which could contain malicious instructions.
Ingestion points: The agent is instructed to read GitHub PR notifications, content from web searches (via Exa), and calendar/email/Slack threads.
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when processing these external data sources.
Capability inventory: The skill possesses the ability to write to the file system, make network requests, control the desktop/browser, and schedule future tasks.
Sanitization: No specific sanitization or filtering logic is provided for the external content before it is processed by the agent loop.

autonomous-agent-harness