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 signed-policy-agentic-commercegit 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/signed-policy-agentic-commerce)<a href="https://agentmods.dev/skills/postoaklabs/ainumbers/signed-policy-agentic-commerce"><img src="https://agentmods.dev/badge/skills/postoaklabs/ainumbers/signed-policy-agentic-commerce/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/signed-policy-agentic-commerce"><img src="https://agentmods.dev/badge/skills/postoaklabs/ainumbers/signed-policy-agentic-commerce.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 | $0.00064 | $0.01714 |
| Opus 5.5 | $0.00026 | $0.00686 |
| Sonnet 5.5 | $0.00013 | $0.00343 |
| Haiku 4.5 | $0.00006 | $0.00171 |
Grade A, and why
signed-policy-agentic-commerce 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 3d 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 — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.
An agent buys under a signed policy and proves the policy governed
Run an agentic commerce chain under a signed Work Mandate, trip the escalation gate on purpose, and replay the gates on the ledger.
Prompt
Act as a procurement agent operating under a signed spend policy. Produce a receipt that proves the policy governed every step. Every server call goes to https://mcp.ainumbers.co/mcp.
The Work Mandate below is synthetic and its signing key comes from a published test seed, so anyone can re-derive the key and re-sign the mandate: take the SHA-256 digest of the ASCII string ainumbers-showcase-mandate-test-key as the 32-byte Ed25519 seed, which yields did:key:z6MknHY2vWeuwJBynEVZEzqfNFuQhBKKUDEyPMdT3Ly9ZVCm. Send the mandate exactly as written, because a single changed byte invalidates the signature. It is valid from 27 September 2026 to 27 September 2028.
MANDATE = {"chaingraph_version":"0.4.0","mandate_type":"work_mandate","tool_id":"work-mandate","generated_at":"2026-09-27T00:00:00Z","policy_parameters":{},"output_payload":{"mandate_type":"work_mandate","scope":{"tool_ids":[],"chains":["agent-commerce-conformance","escalation-sla-supervised-autonomy-receipt"]},"conditions":[],"escalation_triggers":[],"validity":{"not_before":"2026-09-27T00:00:00Z","not_after":"2028-09-27T00:00:00Z"},"principal":{"id":"did:key:z6MknHY2vWeuwJBynEVZEzqfNFuQhBKKUDEyPMdT3Ly9ZVCm"}},"execution_hash":"3c4509eea7acc8c8298cb155ad4ca2a931c2fa4296b57d1cc89d01a92ba1f1cc","audit_signature":{"proof":{"type":"DataIntegrityProof","cryptosuite":"eddsa-jcs-2022","verificationMethod":"did:key:z6MknHY2vWeuwJBynEVZEzqfNFuQhBKKUDEyPMdT3Ly9ZVCm","proofPurpose":"assertionMethod","created":"2026-09-27T00:00:00Z","proofValue":"zfvEY9FAuQLwwMeGqp7a8vpj68aMnYFsRuLrZMfGsfWqLWoC2LZpT7h91qdXoSxjMCckGaZqg3wmVikgMDPZR9o1"}}}
- Call find_chain with "agent commerce conformance" and pick the entry whose chain_name is agent-commerce-conformance. List its four steps from that result and say which ones are callable server-side.
- Call run_chain with chain agent-commerce-conformance, compute "server" and no mandate. Report every step's execution_hash and the composite execution_hash. This is the baseline nobody authorised.
- Call run_chain again on the same chain with compute "server" and mandate MANDATE. Report every step's execution_hash and the composite. Then explain the difference: run_chain verifies the §16 eddsa-jcs-2022 signature, takes mandate_hash from the mandate's own execution_hash, and folds that hash into each step's policy_parameters and into the composite preimage under §22.5 conditional presence. The run without a mandate stays byte-identical to the pre-binding baseline, so the two composites differing is the proof that a policy was in force.
- Recompute the mandated composite yourself. Call verify_execution_hash with policy_parameters taken from that run's composite_artifact.policy_parameters, with output_payload built as { chain, steps } from the same artifact's output_payload, and with claimed_hash set to the composite execution_hash. Leave step_compliance_flags out of output_payload, because §22.8.2 adjacent metadata sits outside the hash preimage. Confirm valid is true and that mandate_hash is one of the policy_parameters you just hashed. Then repeat the step 3 call unchanged and confirm the composite and dedupe.input_hash come back identical.
- Call run_chain with chain escalation-sla-supervised-autonomy-receipt, compute "server", the same MANDATE and escalation_transport "resolve_handle". This chain carries a §21 gate on art-67. Report the run status, the gate decision with the value it read, which step was skipped, and the escalation record with its record_hash. Confirm the record carries the same mandate_hash as the step receipts, so the open item names the policy it was opened under.
- Change one character inside MANDATE's scope, such as the last letter of the first chain name, and call run_chain on agent-commerce-conformance again. Quote the error verbatim and say at which point the run stopped.
- Open the ledger_url from the mandated commerce run in step 3. Report what the §21 gate replay shows and whether the §16 signature check and the mandate_hash binding pass in the browser.
- Now show the counterparty would accept the agent: on https://ainumbers.co/chaingraph/art-129-webbotauth-signature-verifier.html, use the page's WebMCP tool on the page's own synthetic sample to verify a WebBotAuth signed request header. Record the execution_hash and confirm DevTools shows no request leaving the tab.
- Call build_session_receipt over four hashes in call order: the unmandated composite, the mandated composite, the escalated chain's composite, and the in-page identity hash. Anchor the root with anchor_hash on https://anchor.ainumbers.co/mcp using the FreeTSA RFC 3161 authority. Return the session root, the anchor receipt, and a five-line summary covering who authorised the work, which code ran, where the gate halted it, the mandate hash that binds all of it, and what a person still has to close out through the open resolve handle. 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.
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.
- 3d ago Changed 6298e8109230
- 1mo ago Changed · +5 lines f21d4a816522
- 1mo ago First seen · 49 lines · 64 tokens per session scan A 49b47268d723
signed-policy-agentic-commerce is a skill published in the GitHub repository PostOakLabs/ainumbers (12 stars, last pushed today), licensed MIT. It adds 64 tokens to every session and 1,714 once invoked, about $0.0003 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.
signed-policy-agentic-commerce compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| signed-policy-agentic-commercePostOakLabs/ainumbers | Skill | 12 | today | MIT | A | 64 | Claude Code, Codex |
| tbank-grocery-ordericyberdeveloper/tbank-mcp | Skill | 57 | 3d ago | MIT | A | 124 | Claude Code |
| tbank-ticketsicyberdeveloper/tbank-mcp | Skill | 57 | 3d ago | MIT | A | 155 | Claude Code |
| shop-checkwaldo-van-der-code/deutsches-recht-mit-claude | Skill | 45 | 5d ago | MIT | A | 137 | Claude Code, Codex |
| recht-verbraucher-momarcode1ThomasMoreAI/legal-skills-open | Skill | 136 | 1mo ago | Apache-2.0 | A | 73 | Claude Code, Codex |
| common-law-ucc-sales-goodsThomasMoreAI/legal-skills-open | Skill | 136 | 1mo ago | Apache-2.0 | A | 127 | Claude Code, Codex |
| paymob-integrationPaymobAccept/Paymob-AI-Integration-Skill | Skill | 36 | 14d ago | MIT | A | 52 | Claude Code, Codex |