lcx-contribute-bug-fix

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The required workflow materializes and syncs runtime sources by cloning public GitHub repositories (gh repo clone / git clone for code-yeongyu/lazycodex and openai/codex) and then runs commands/tests in those checkouts, so outsider-authored repository text (code/docs/issues/tests) is ingested into the agent’s LLM context via the debugging/investigation steps.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly clones and uses remote repositories at runtime (e.g., gh/git clone of code-yeongyu/lazycodex and openai/codex via https://github.com/code-yeongyu/lazycodex and https://github.com/openai/codex), and those fetched sources are then run/tested as part of the fix workflow, so remote content can execute code or control the agent's behavior.