GitHub Watch

This skill appears to do what it says, but consider the following before installing: - Protect your GitHub token file: it is read as plaintext from ~/.openclaw/secrets/github_token (optional) — keep file permissions restricted and only supply a token with the minimal scope you need. - Review and trust mail-client and nextcloud-files skills: this skill imports/executes the mail-client script and runs the nextcloud script to deliver content, so those other skills' code will execute on your machine — if they are untrusted, they could send data or perform unexpected actions. - Local data writes: the seen-store and config live under ~/.openclaw; this is expected but you should verify that path if you have strict data controls. - Prompt-injection handling: the skill includes an explicit untrusted wrapper to reduce prompt injection risk when feeding external repo descriptions to an LLM, but wrapping is a textual mitigation only — continue to treat model outputs cautiously and avoid automatic execution of any instructions found in repo text. - If you do not want the agent to autonomously run the pipeline, keep the skill user-invocable and control cron/automation settings; the skill itself does not set always:true. Overall, the footprint is proportionate to its stated purpose; review the delivery skills and your token handling policy before use.