← ClaudeAtlas

claude-rescuelisted

Ask Claude Code for rescue diagnosis or explicit write-mode repair when Codex is stuck, failing tests, or uncertain about recovery.
yilibinbin/external-models-for-codex · ★ 1 · AI & Automation · score 64
Install: claude install-skill yilibinbin/external-models-for-codex
# Claude Rescue Use this skill when Codex needs an independent Claude diagnosis for a stuck implementation, failing validation, confusing git state, or unclear next recovery step. Run: ```bash node "${CODEX_PLUGIN_ROOT}/scripts/claude-companion.mjs" rescue "$ARGUMENTS" ``` ## Natural-Language Claude Routing <!-- routing:rescue routing:write-mode-explicit --> - Do not ask the user to write `--quality`, `--model`, or `--effort` unless troubleshooting the plugin itself. - Default to `--quality auto` for manual Claude review, plan, rescue, and multi-review unless the command documents a stricter default. - Use `--quality strong` for deep, strict, high-risk, migration, release, or difficult diagnosis/planning requests. - Use `--quality max` only when the user explicitly asks for the strongest local Claude pass. - If the user names a concrete Claude model or effort, pass it as explicit argv tokens outside quoted `$ARGUMENTS`. - Keep rescue read-only unless the user explicitly asks Claude to write or repair files. - Route explicit repair requests to `rescue --write`; inspect the resulting git diff before reporting success. - Before choosing foreground for an unforced review, run `recommend-execution-mode --json` or apply the same threshold: foreground only for tiny one-to-two-file work; use background for untracked directories, more than two files, more than roughly fifty changed lines, multi-role/adversarial/rescue work, or unclear scope. - Do not substitute strong local Clau