claude-like-codex-code-reviewlisted

# Claude-like Codex Code Review Use this invocation contract: ```text $claude-like-codex-code-review [low|medium|high|xhigh|max] [--fix] [--comment] [<pr#>|<branch>|<path>] ``` All arguments are optional and may appear in any order. Parse them before reviewing: - The first effort token selects the review profile. Default to `high` when omitted. - `--fix` authorizes applying retained findings to the local working tree after the review. It does not authorize commits, pushes, branches, PR creation, or unrelated refactors. - `--comment` authorizes publishing the retained findings to one resolved GitHub PR. It does not authorize approving the PR, requesting changes, or posting elsewhere. - Accept at most one target: `123` or `#123` for a PR, a resolvable Git branch/ref, or an existing file or directory. Preserve quoted paths containing spaces. - Reject unknown flags, multiple effort levels, and multiple targets with one concise correction. Do not silently reinterpret them as review focus text. Resolve an ambiguous target in this order: explicit PR number, explicit path beginning with `./`, existing path, then resolvable Git ref. The user may still provide ordinary-language focus instructions after invoking the skill; treat those as review criteria rather than positional targets when they do not match the grammar above. Follow this sequence: parse arguments → fix scope and code version → collect rules → summarize changes → run the selected discovery profile → independently ve