The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the way it handles and routes untrusted data payloads.
Ingestion points: Data enters the system via the --payload CLI argument, the payload JSON field in API requests, and external config.json files.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore potential commands embedded within the input payloads.
Capability inventory: The skill possesses the ability to execute CLI commands through the openclaw tool and initiate network requests via the Python requests library.
Sanitization: There is no evidence of input validation or sanitization before data is passed to other instances or processed by the system.
[EXTERNAL_DOWNLOADS]: The skill documentation mandates the installation of an external software package.
Evidence: Users are instructed to perform pip install openclaw-sdk to enable skill functionality.
[COMMAND_EXECUTION]: The skill relies on local command-line tools to perform its orchestration and delegation tasks.
Evidence: The documentation features several examples of using the openclaw orchestrator CLI for routing, fanout, and session management.

orchestrator