The Agent Skills Directory

[COMMAND_EXECUTION]: Describes the built-in Cmd tool in references/builtin-tools.md which allows local shell command execution with security controls such as allowed_cmd_prefixes and allowed_workdir_roots.
[EXTERNAL_DOWNLOADS]: References the Agently Python framework and third-party libraries like playwright and pyautogui in references/builtin-tools.md. All source links in references/source-map.md point to the official AgentEra GitHub organization.
[DATA_EXFILTRATION]: Documents Search and Browse tools in references/builtin-tools.md that perform network operations for information retrieval as part of their documented features.
[PROMPT_INJECTION]: Identifies an indirect prompt injection surface in references/builtin-tools.md. 1. Ingestion points: Search and Browse tools retrieve external data. 2. Boundary markers: Not explicitly mentioned for tool result injection. 3. Capability inventory: Includes shell execution via Cmd and network access. 4. Sanitization: Uses allowed_cmd_prefixes for command validation.

agently-tools