القائمة
Monitor Amp Free quota and handoff before exhaustion. Use when user says quota-watch, check quota, or monitor usage.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Plan and consume execution metadata for bd/beads issues. Use when Codex inspects, triages, claims, tackles, or delegates beads; assign or read suggested agent type, model, reasoning effort, execution mode, and parallel group using the documented execution_* metadata keys.
Set up a private GitHub Dolt repo for personal beads and configure sync for the current project
Initialize or use beads in stealth mode for a project that does not use beads, with issue data stored in a separate syncable repository. Use when the user says "init with stealth", "stealth beads", "use beads without touching this repo", or asks for external BEADS_DIR setup.
| name | quota-watch |
| description | Monitor Amp Free quota and handoff before exhaustion. Use when user says quota-watch, check quota, or monitor usage. |
Periodically check Amp Free tier quota. Warn when low. Handoff before exhaustion.
Run scripts/check-quota.sh immediately. Output is: remaining total percent.
After approximately every 15 tool calls, run:
.claude/skills/quota-watch/scripts/check-quota.sh
Do NOT mention the check to the user unless a threshold is hit. Keep monitoring silent.
Parse the three numbers from the script output: remaining total percent.
| Percent | Action |
|---|---|
| ≤ 10% | Warn user: ⚠ Quota at {pct}% (${remaining}/${total}). Wrapping up current task. Then finish only the immediate task. |
| ≤ 5% | Stop all new work. Execute Emergency Handoff below immediately. |
When ≤ 5%, execute these steps in order — skip any that don't apply:
git add only files you changed, then git commit -m "wip: quota handoff"git pushbd update <id> --status in_progresshandoff tool with a goal summarizing: what was done, what remains, current branch, any failing tests.Keep the handoff goal under 3 sentences.
Each check costs ~1 tool call (~30 output tokens). At 15-call cadence that's ~7% overhead. This is acceptable for sessions > 50 tool calls. For quick tasks (< 20 calls), checking once on load is sufficient.