remote-ssh-bridge

Standard SSH command templates for a remote operator machine (bird reads, Puppeteer runs, inbox-style messaging).

This skill is a small collection of SSH command templates and appears to do what it says, but there are practical problems you should fix before using it: (1) The scripts expect REMOTE_TARGET but the skill metadata doesn't declare it — set REMOTE_TARGET explicitly (e.g., export REMOTE_TARGET="user@host") and verify it points to the correct host. (2) The message script interpolates MESSAGE into a remote shell command without escaping; if the agent or other automation supplies message text, a crafted message could execute arbitrary commands on the remote host. Mitigations: sanitize/escape message input or pass the message via stdin (e.g., echo "$MESSAGE" | ssh "$REMOTE_TARGET" 'some-safe-handler'), or use printf '%s' with proper quoting on the remote side. (3) Replace the placeholder defaults (neill@<YOUR_REMOTE_HOST>) with no default or a clearly documented required variable to avoid accidental connections. (4) Only run this skill when you have reviewed the exact remote command placeholders and you trust the source of any messages passed to it. If you plan to allow autonomous agent invocation, restrict or validate message inputs to avoid accidental remote command execution.

Type: OpenClaw Skill Name: remote-ssh-bridge Version: 0.1.3 This skill bundle provides basic SSH command templates for interacting with a remote host, as explicitly stated in its purpose. The scripts (`check-sapconet.sh`, `msg-sapconet.sh`) contain clear `TODO` placeholders, requiring the user to configure the `REMOTE_TARGET` and replace placeholder commands. While the direct interpolation of the `MESSAGE` variable in `msg-sapconet.sh` could be a shell injection vulnerability if the `TODO` were replaced with an executing command without proper sanitization, in its current form, the message is safely enclosed within single quotes inside an `echo` command on the remote host. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts against the AI agent in the `SKILL.md`.