Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/PostOakLabs/ainumbersWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/postoaklabs/ainumbers/air-gapped-helm-handover)<a href="https://agentmods.dev/commands/postoaklabs/ainumbers/air-gapped-helm-handover"><img src="https://agentmods.dev/badge/commands/postoaklabs/ainumbers/air-gapped-helm-handover/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/postoaklabs/ainumbers/air-gapped-helm-handover"><img src="https://agentmods.dev/badge/commands/postoaklabs/ainumbers/air-gapped-helm-handover.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00032 | $0.00442 |
| Opus 5 | $0.00016 | $0.00221 |
| Sonnet 5 | $0.00006 | $0.00088 |
| Haiku 4.5 | $0.00003 | $0.00044 |
Grade A, and why
air-gapped-helm-handover scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Air-gapped control plane: the agent runs it, the human releases it, the bundle verifies without us
You are connected to a local Helm daemon at 127.0.0.1:4173. Run a CECL allowance workflow end to end, prove it, and get the evidence out the only way the daemon allows.
- Call catalog.search with "cecl". Pick the quarterly allowance pack. Call workflow.describe and workflow.manifest_get; list the nodes, the gates, and the manifest digest.
- Call workflow.dry_run on it (declare the io.modelcontextprotocol/tasks extension). Poll tasks/get until status is completed. Report the execution_hash.
- Call workflow.run on the same pack. Poll to completion. Report the execution_hash and whether it equals the dry run's.
- Call artifact.get and then artifact.verify on the real run. Report the per-step digests and the replay verdict.
- Attempt evidence.export with no ticket. Quote the error verbatim. Then tell me exactly what the human at the Helm UI has to do to mint a consent ticket, and wait for me to paste it.
- Once I paste the ticket, call evidence.export with it. Save the digest-level record.
- Cross-check the law: find the same kernel on mcp.ainumbers.co (find_tool "CECL allowance"), call it with the pack's sample inputs, and compare the kernel_digest and execution_hash to the Helm run. State whether the local daemon and the public worker agree.
- Call anchor_hash on the Helm run's execution_hash (Sigstore TSA). Return the receipt.
- Write the handover: what ran locally, what never left the machine, where the human consent sat in the flow, and how an auditor verifies the bundle offline with
helmd verify <bundle> --keys <publicKeys.json>and no network.
Finish by stating the execution_hash you received and the ledger link a human can verify it at.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 22 lines · 32 tokens per session scan A 4055d32f7a62
air-gapped-helm-handover is a command published in the GitHub repository PostOakLabs/ainumbers (5 stars, last pushed today), licensed MIT. It adds 32 tokens to every session and 442 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-06.
Other commands, from other repositories
risk
Risk reassessment with updated due diligence.
screen
Comprehensive sanctions and PEP screening.
monitor
AML transaction monitoring and suspicious activity reporting.
onboard
Customer onboarding with Step 0 verification (HITL consent required).
onboard-interactive
Interactive customer onboarding with Step 0 independent verification (dialog-based, 17 stagegates).
refresh
Periodic customer review with independent verification.