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.
npx skills add PostOakLabs/ainumbers --skill agent-spend-policy-stressgit 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/skills/postoaklabs/ainumbers/agent-spend-policy-stress)<a href="https://agentmods.dev/skills/postoaklabs/ainumbers/agent-spend-policy-stress"><img src="https://agentmods.dev/badge/skills/postoaklabs/ainumbers/agent-spend-policy-stress/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/skills/postoaklabs/ainumbers/agent-spend-policy-stress"><img src="https://agentmods.dev/badge/skills/postoaklabs/ainumbers/agent-spend-policy-stress.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 | $0.00046 | $0.00625 |
| Opus 5.5 | $0.00018 | $0.00250 |
| Sonnet 5.5 | $0.00009 | $0.00125 |
| Haiku 4.5 | $0.00005 | $0.00063 |
Grade A, and why
agent-spend-policy-stress 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 15d 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.
How it starts
The opening of the file, as written. The whole thing — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent spend policy stress
Stress my agent’s spend policy against 10,000 synthetic transactions and export the Policy Mandate.
Prompt
Find out what your agent's spend policy actually does before a live agent finds out for you.
- Write down the policy as you believe it is (per-transaction cap, daily cap, merchant allowlist). This is the claim you are testing.
- Call simulate_spend_policy with 10,000 synthetic transactions spanning: under-cap, at-cap, over-cap, burst-across-midnight, and allowlist-boundary cases. Record the approve/decline counts per class.
- Call simulate_agent_spend_policy for the agent-shaped variants (parallel requests, retry storms, currency round-trips). Record where the policy and the agent disagree.
- Find the worst class: the transactions where the outcome contradicts your step-1 claim. Quote three examples with inputs.
- Adjust the policy (or file the fix), re-run only the failing class, and quote the new counts.
- Export the effective Policy Mandate your simulation ran under; record its digest.
- build_session_receipt over the runs; give me the ledger link.
- Write the note for treasury: what the policy really enforces, the three example transactions, the mandate digest, and the ledger link. Call shape: every ChainGraph node tool takes its arguments nested under one wrapper object, e.g. {"policy_parameters": { ... }}; flat arguments are discarded by schema validation. If a named tool is not in your tool list, call call_tool with { "name": "", "arguments": { … } }. It runs the same validation and returns the same receipt. GPU note: the simulate_spend_policy tool computes in your browser, so the MCP endpoint returns no execution_hash; export the Policy Mandate artifact the page produces and pass that full artifact to verify_execution_hash as claimed_hash.
How to run
Connect an MCP-capable host to https://mcp.ainumbers.co/mcp, then give the prompt above to the assistant.
Use synthetic inputs only. Never paste personal or production data.
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.
- 15d ago Changed f49c8b6f9c25
- 19d ago First seen · 42 lines · 46 tokens per session scan A 16bdbbed5194
agent-spend-policy-stress is a skill published in the GitHub repository PostOakLabs/ainumbers (11 stars, last pushed yesterday), licensed MIT. It adds 46 tokens to every session and 625 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-27.
Other skills, from other repositories
risk-assessment
Deterministic heuristic risk scoring using a four-factor weighted model (Geographic 30%, Customer 35%, Product 25%, Channel 10%) aligned with Basel III, EBA Guidelines, FATF standards, and FinCEN guidance. Bands: LOW (0-20), MEDIUM (21-60), HIGH (61-80), CRITICAL (81-100).
transaction-monitoring
AML transaction monitoring for ongoing customer relationships. Identifies suspicious patterns including structuring, layering, and unusual activity. Assesses against regulatory thresholds and prepares Suspicious Activity Reports (SARs) with HITL decision gates.
prompt-proximity-architecture
Turn an approved measurement charter, ICPs, and buyer jobs into a budget-aware prompt coverage blueprint across proximity bands, aided status, information acts, journey states, roles, locales, evidence grades, partitions, and measurement lanes. Use before prompt wording to define required, optional, and prohibited…
backtesting-frameworks
Build robust backtesting systems for trading strategies with proper handling of look-ahead bias, survivorship bias, and transaction costs. Use when developing trading algorithms, validating strategies, or building backtesting infrastructure.
japan-bav-und-corporate-pension-iorp
Japanisches betriebliches Altersversorgungssystem und IORP-Vergleich für europaeische Konzerne. Normen: IORP-II, japanisches Pensionsrecht DB-Pensions-Act. Prüfraster: Leistungsunterschiede, Finanzierungsanforderungen, Konvergenz. Output: Vergleichsanalyse Japan-bAV vs. EU-IORP. Abgrenzung: nicht Expatriate-Planung.
au-tax-research-ryanduguid
Use when researching an Australian tax question for a file note or draft advice: frame the question, verify each primary source and its status for the period, record alternative views and write a review-ready research file note.