Rails CI Fixer

Autonomously fix failing CI on Rails PRs using a tiered escalation loop. Use this skill whenever a Rails pull request has failing CI — RSpec failures, RuboCo...

This skill appears to do what it says, but it runs repository tests and can execute arbitrary code from the repo. Only use it on repositories you own or fully trust. If you install or run it: 1) use a fine-grained GH personal access token scoped to just the repo (actions: read, contents: write), set an expiration, and rotate/revoke after use; 2) run the agent in an isolated user account, container, or VM when operating on unfamiliar code; 3) enforce branch protection on main and review diffs before merging (the skill pushes only to feature branches, but you should still review automated commits); 4) note the small metadata inconsistency (skill declares no primary credential in registry metadata but requires GH_TOKEN in runtime instructions) and confirm the platform will prompt for the GH_TOKEN appropriately. If you want higher assurance, request an explicit listing of required runtimes (rspec usage is via bundle exec) and confirmation that the agent will not accept run_id or other inputs from untrusted external sources.

Type: OpenClaw Skill Name: rails-ci-fixer Version: 1.1.4 The skill automates Rails CI fixes by executing shell commands, running test suites (bundle exec rspec), and pushing code to GitHub repositories. While these actions are aligned with its stated purpose and the documentation (SKILL.md, references/security.md) includes explicit security warnings and mitigations—such as treating CI logs as untrusted data to prevent prompt injection—the inherent risk of arbitrary code execution and repository write access via an AI agent warrants a suspicious classification. No evidence of intentional malice, data exfiltration, or obfuscation was found.