fix(ci): resolve critical security issues in GitHub Actions workflows

[h0lybyte]

Summary

• Disable atlas process workflow — i-atlas-process.yml passed attacker-controlled issue bodies through an LLM, then injected unvalidated LLM output directly into shell commands via ${{ }} interpolation. Changed trigger from issues:opened to workflow_dispatch to disable it. Atlas processing should move to a dedicated external application.
• Fix command injection in shell actions — kbve-shell/action.yml, utils-astro-deployment.yml, and utils-nx-kbve-shell.yml all interpolated ${{ inputs.* }} directly into run: blocks, allowing shell metacharacter injection. Moved inputs into env: variables so the shell treats them as data, not code.
• Enable Trivy CVE blocking — utils-publish-docker-image.yml had exit-code: 0 on the Trivy scan, meaning CRITICAL/HIGH vulnerabilities were logged but never blocked publishing. Changed to exit-code: 1.
• Harden i-label.yml — Added top-level permissions: {issues: write, contents: read} (was inheriting default write-all) and timeout-minutes: 5 to all 16 jobs. Fixed misleading [Help] comment on assign_staging.

Test plan

• Label automation still works (apply a label to an issue, verify project migration runs)
• Astro deployments still build (trigger via ci-main)
• Docker publish pipeline blocks on CRITICAL CVEs instead of silently passing
• Atlas workflow no longer triggers on new issues

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

• .github/workflows/i-atlas-process.yml