Schedule After Usage Reset

Automatically find the usage reset time and call /schedule with it. No questions asked.

Overview

When the user wants to defer a task until after their Claude usage/token limit

resets, this skill removes the manual step of looking up the reset time. It reads

the reset timestamp from the Anthropic usage API, converts it to the user's local

timezone, adds a small buffer, and hands the task off to the /schedule skill at

that exact time. The user never has to figure out when the limit lifts — they

just say what they want run.

Prerequisites

• Running inside Claude Code with a valid OAuth credential stored under the

Claude Code-credentials keychain entry (read via /usr/bin/security).

• Network access to https://api.anthropic.com to query the usage endpoint.
• python3 available on PATH (used to parse the JSON token and API response).
• The schedule skill installed — this skill defers the actual scheduling to it.

Instructions

1. Get the reset time

Fetch from the Anthropic usage API:

token_json=$(/usr/bin/security find-generic-password -s 'Claude Code-credentials' -w 2>/dev/null)
access_token=$(echo "$token_json" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d.get('claudeAiOauth',{}).get('accessToken',''))")
curl -s -H "Authorization: Bearer $access_token" \
     -H "anthropic-beta: oauth-2025-04-20" \
     -H "User-Agent: claude-code/2.1" \
     "https://api.anthropic.com/api/oauth/usage"

Parse fivehour.resetsat (or sevenday.resetsat). Convert UTC → user's local

timezone. Add 5 minutes as buffer.

2. Get the task

If not provided as an argument, ask: "What should I run after the reset?"

3. Call /schedule

Invoke the schedule skill, passing the task and the reset time + 5min buffer:

/schedule "<task>" at <HH:MM> <timezone>

The schedule skill handles everything from there — just like calling it directly.

Output

• A /schedule invocation has been created for the user's task at the reset time

plus the 5-minute buffer, expressed in the user's local timezone.

• A short confirmation of what was queued and when it will run (for example:

"Queued to run at 14:35 America/New_York, just after your usage resets").

• The /schedule skill owns execution from that point — this skill's job ends once

the handoff has been made.

Error Handling

• Rate limited (HTTP 429): wait 15 seconds and retry the usage request once.
• Still failing after the retry: ask the user for the reset time directly and

proceed with /schedule using their answer.

• Missing or empty access token: the keychain lookup returned nothing — ask the

user for the reset time rather than guessing.

Examples

Defer a build until the limit lifts

User: schedule a full test run for after my usage resets
→ Read five_hour.resets_at from the usage API (e.g. 19:30 UTC)
→ Convert to local time and add 5 min → 14:35 America/New_York
→ /schedule "run the full test suite" at 14:35 America/New_York

Reset time unavailable

User: queue this report for when my tokens refresh
→ Usage API still 429 after one retry
→ Ask: "What time does your usage reset?" → user says 3:00 PM
→ /schedule "<task>" at 15:05 <timezone>

Resources

• Anthropic usage endpoint: https://api.anthropic.com/api/oauth/usage

(fivehour.resetsat / sevenday.resetsat carry the reset timestamps).

• The schedule skill — this skill is a thin front-end that resolves the reset

time and then defers all scheduling to it.