The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/team-handoff.sh script executes several external utilities including cloudflared for authentication, python3 for JSON manipulation, ssh for remote access, and the openclaw CLI for gateway operations.
[PRIVILEGE_ESCALATION]: The skill includes an 'SSH operator fallback' feature that executes commands on a remote host using sudo -n -u $REMOTE_USER. This allows the script to acquire the privileges of the specified remote user to run the gateway CLI.
[DYNAMIC_EXECUTION]: The script dynamically generates a configuration file (openclaw.json) for the OpenClaw CLI. This configuration utilizes an exec secret provider, which instructs the CLI to execute the cloudflared binary at runtime to obtain identity tokens (Cf-Access-Token).
[INDIRECT_PROMPT_INJECTION]: The skill processes a handoff file provided by the user and sends its entire content to a remote agent, creating a surface for indirect prompt injection if the file contains untrusted data.
Ingestion points: The content of the file path provided to the --message-file argument in scripts/team-handoff.sh.
Boundary markers: None identified; the message content is passed directly into the JSON payload for session creation.
Capability inventory: The skill has the capability to make network requests to a Gateway and execute commands on a remote host via SSH.
Sanitization: No sanitization or escaping is performed on the handoff message content before it is sent to the remote agent.

team-handoff