Codex Reset Credits

Overview

Query Codex banked reset credits through the bundled read-only script. Return a concise, sanitized answer with available count and expiration times.

Workflow

1. Run the bundled script:

   python3 "$SKILL_DIR/scripts/query_reset_credits.py" --json
   

   If $SKILL_DIR is not available, locate this Skill folder by the directory containing this SKILL.md, then run scripts/query_reset_credits.py inside it.

2. Parse the sanitized JSON result.

3. Answer in the user's language using this format:

   查到了。你当前有 `<available_count>` 个可用 Codex 重置机会。
   
   最早到期：`<earliest expires_at.time_local>`
   
   | # | 状态 | 类型 | 到期时间 | UTC |
   | --- | --- | --- | --- | --- |
   | 1 | `available` | `codex_rate_limits` | 2026-07-16 15:51:44 CST | 2026-07-16 07:51:44 UTC |
   

   Keep the answer concise. Do not wrap the table in a code block. Mention that output is sanitized only when useful or when sharing with someone else.

Safety Rules

• Do not ask the user to paste access tokens.
• Do not print access tokens, full auth.json, full account ids, full credit ids, or raw backend responses.
• Do not write real backend responses to the workspace.
• Do not redeem reset credits or call mutation/control endpoints.
• Treat HTTP 401/403 as an auth/login problem, not proof that no reset credits exist.

Failure Handling

If the script returns ok: false, report the error and next step:

• missing_auth: open Codex Desktop and log in.
• invalid_auth_shape: Codex auth storage changed or is unreadable by the script.
• missing_access_token: refresh or relogin Codex Desktop.
• http_error with 401/403: relogin Codex Desktop.
• http_error with 429: wait for retry_after when present.
• invalid_json: backend response shape changed; inspect endpoint behavior without exposing secrets.

Link Runbook

When the user provides a link to USE_WITH_CODEX.md, follow that runbook. It exists for users who have not installed this Skill yet.

Maintenance Reference

Read references/api.md only when debugging the endpoint, auth headers, or response-field mapping.