The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md require the agent to execute a local Python script located at ~/Projects/agent-scripts/skills/openclaw-relay/scripts/openclaw_relay.py using python3. It utilizes various subcommands like targets, resolve, and ask to facilitate the relay.
[DATA_EXFILTRATION]: The skill is designed to transmit messages to an external Discord session. While this is the intended functionality of the OpenClaw relay, it represents a network egress channel where agent context or user data is sent to a third-party service (Discord) that is not part of the standard whitelisted domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes data returned from the Discord relay, creating a vulnerability to malicious instructions embedded in Discord messages.
Ingestion points: Data enters the agent's context through the output of the python3 ... ask command, which retrieves replies from the Discord-backed agent.
Boundary markers: None are present. The instructions do not specify delimiters or warnings for the agent to ignore potentially malicious instructions within the relay's output.
Capability inventory: The skill maintains shell execution capabilities (python3) and network access (via the relay script).
Sanitization: No sanitization, filtering, or validation logic is defined for the content received from the Discord relay.

discord-clawd