June 15 Readiness Audit — Claude Code Pre-Cliff Assessment

10 questions · ~5 minutes · browser-only, no signup, no telemetry · CC0 · v1.0 (2026-05-24)

The cliff: On 2026-06-15 (UTC), Anthropic splits programmatic billing. claude -p and Claude Agent SDK invocations route to a separate credit bucket (Pool 2). Documentation: docs.claude.com/en/api/billing. This audit estimates how much of your current setup is exposed to the change.

Disclosure: the recommendations below cite the cc-guard ($19, released 2026-05-22) where its chapters address a specific gap. The free Chapter 1 preview Gist and the free Trigger 15 chapter Gist may already be enough; the book is only useful if multiple gaps below score high.

Audit

1. How often do you invoke claude -p (one-shot) or Claude Agent SDK?

Includes cron jobs, CI runners, automated agents.

Never. Only interactive claude REPL. Rarely. 1–5 times/month. Sometimes. 5–50 times/month. Frequently. 50+ times/month or cron-scheduled.

2. What is your current monthly Claude Code spend?

Combine subscription plan cost (Pro/Max) and any overage API charges.

Under $20 (Pro plan, no overage). $20–$100 (Pro with overage or Max 5x). $100–$200 (Max 20x). Over $200 (Max + API overage).

3. Do you have a /cost baseline from before 2026-06-14?

A snapshot of claude /cost --json output you can compare against post-cliff.

Yes, saved to file. No, but I plan to capture one before 2026-06-15. No, and I haven't planned for one.

4. Do you have a cost-guard hook on PreToolUse to detect runaway burns?

A hook that aborts or warns when daily token consumption exceeds a threshold.

Yes, installed and tested. Aware of the concept, not installed. Not familiar with the pattern.

5. How many sub-agents (.claude/agents/*.md) does your project define?

Sub-agent invocations from headless parent contexts may route to Pool 2.

None. 1–3. 4–10. 10+ or heavy parallel dispatch.

6. How many high-frequency Skills (SKILL.md with broad triggers) does your setup load?

Skills that activate on most user prompts compound the cache_creation cost.

None or only narrow-trigger Skills. 1–3 broad-trigger Skills. 4–7 broad-trigger Skills. 8+ broad-trigger Skills.

7. Have you verified your CLAUDE.md is under 5,000 tokens per turn?

Bloated CLAUDE.md compounds with every session, especially under Pool 2 calculations.

Yes, measured and below 5,000 tokens. Roughly checked, probably fine. Never measured.

8. Do you know which alternative AI providers cover your top 3 use cases?

If Pool 2 charges spike post-cliff, can you switch a workload to Codex / DeepSeek / local LLM?

Yes, I have tested alternatives for my top use cases. Aware of alternatives, never tested. Have not considered alternatives.

9. Have you set a billing alert in your Anthropic account?

Independent of in-app /cost — the actual Anthropic billing dashboard alert.

Yes, alerts configured. Aware of the feature, not configured. Not familiar.

10. Do you have a refund/escalation template ready for unexpected charges?

A pre-written Issue body that captures plan tier, /cost JSON snapshots, and the specific cap breach pattern.

Yes, template prepared. Know the pattern, no template. Would not know where to start.

out of 100 — Higher score means LESS prepared (more exposure to the cliff)

Your top gaps

The full migration framework — including Pool 1/Pool 2 attribution Chapter 7, cache_creation Chapter 8, Path B' (Codex) Chapter 5, and Path D (local LLM) Chapter 6 — is in cc-guard ($19, 241 pages, released 2026-05-22). Free Chapter 1 preview and free Trigger 15 chapter may be enough — the book is only useful if multiple gaps above score high.