When a third-party skill of Hermes Agent is intercepted by a security scan, don't add --force as soon as you react. Use hermes skills inspect to see the source, permissions, scripts, and scan results first. Only consider forced installation if it is not at a dangerous level and you can confirm the reason.
What is the scan protecting against?
The official skills document mentions that the skills installed on the Hub will check for risks such as data leakage, prompt injection, disruptive commands, and supply chain signals. In other words, it is not "troublesome", but it is blocking skills that may read keys, upload files, or execute dangerous commands.
Secure processing sequence
- Let's look at the skill source first: the trust level of official, trusted repositories, and community repositories is different.
- Run inspect: Confirm the repo, detail page, installation command, audit status, and script content.
- Look at the interception level: warn/caution class can be judged manually, dangerous should not be bypassed.
- Use
--forceonly when you understand the risks.
What situations should not be pretended
- The skill requires reading the entire home directory or key directory, but the feature does not require it.
- Install the script to download the unknown binary and execute it immediately.
- The README statement is clearly inconsistent with the actual script behavior.
- The source repository is new, has no maintenance record, and has no credible citations.
Third-party skills can make Hermes strong, but it is also equivalent to handing over new tools to the agent. The correct posture is to audit first, then install, and then test in a low-privilege environment before using it for real projects.
