Safely search and review Clawhub skills by keyword, showing details and risk before asking for explicit approval to install.

This skill appears to be a simple wrapper around the 'clawhub' CLI to show top search results, but the documentation (SKILL.md) is incomplete and self-contradictory. Before installing or enabling it, ask the author to: 1) remove placeholder text and clearly document the exact commands used; 2) reconcile the sandbox/no-write/no-exec claims with the script (either make it truly read-only or remove the claim); 3) implement and show the approval + install flow if the skill will perform installs (currently it only prompts but does not execute installs); 4) make logging explicit and configurable (current script appends to logs/clawhub-search.log); and 5) fix parsing bugs (e.g., download counts with commas will break numeric comparisons). If you still want to test it, run it in an isolated environment where creating 'logs/' and executing the 'clawhub' CLI are safe, and inspect the output and any created files. Because of the mismatches and missing pieces, treat this skill as untrusted until the author fixes the documentation and implementation.

Type: OpenClaw Skill Name: clawhub-search-verify Version: 1.0.3 The skill is classified as suspicious due to significant discrepancies between its `SKILL.md` documentation and the `cli-wrapper.sh` implementation. The `SKILL.md` explicitly instructs the AI agent to "run `clawhub install <slug>` on approval," which is not implemented in the provided shell script, creating a high risk of prompt injection where the agent might execute arbitrary installations. Furthermore, `SKILL.md` falsely claims "no filesystem write" capability, directly contradicted by `cli-wrapper.sh` writing user search terms to `logs/clawhub-search.log`, misrepresenting the skill's actual permissions and behavior.