algovoi-spend-guardrail-lite 0.1.2

Open, content-addressed pre-payment ALLOW/DENY decisions bound to the pinned AlgoVoi substrate (agent + mandate + policy; L1 frozen)

algovoi-spend-guardrail-lite

The open, lite origination layer for AlgoVoi Spend Guardrail. One call an agent platform makes before executing a payment: bind a categorical ALLOW / DENY decision to the agent it was made for (agent_ref), the spend authority it was checked against (mandate_ref), and the policy in force (policy_bound_ref), each imported by hash, into a deterministic, recomputable guardrail_ref. Moves no funds.

The whole pre-payment decision becomes one content address. Swap the agent, the mandate, the policy, or the verdict and the guardrail_ref diverges: an ALLOW made under one policy snapshot does not recompute under a rotated policy. Additive over the frozen substrate (Apache-2.0):

guardrail_ref = "sha256:" + SHA-256(JCS({agent_ref, mandate_ref, policy_bound_ref, verdict}))

Lite vs commercial

	Lite (this package)	Spend Guardrail
Licence	Apache-2.0, open	Commercial OEM
Decision binding	content-addressed guardrail_ref	same, Falcon-1024 signed receipt
Enforcement	bring your own verdict	full Agent Passport + Payment Mandate stack
Verifier	recompute offline	maintained verifier

Pin lite, anchor a vector hash, carry the NOTICE → free v0 key; upgrade for post-quantum signing + the enforcement stack.

Install & use

pip install algovoi-spend-guardrail-lite

from algovoi_spend_guardrail_lite import guardrail_ref, verify

# each input is imported by hash: a passport_ref, a mandate_ref, and a policy_bound_ref
agent  = "sha256:b3594e33998af01bd1ad208172c5c1ac586daa8c75781379f034d97e50b1a9be"
mandate = "sha256:a4f8cb5ee09b29478ac1cc2f468d66e16d3d25f7a229a31d22ad521e11d04d35"
policy = "sha256:aaee2091799f376ee8cac802ea4920feaa4eca52950488a3e047ff82e6959a21"

ref = guardrail_ref("ALLOW", agent, mandate, policy)   # bind the decision
verify(ref, "ALLOW", agent, mandate, policy)           # True
verify(ref, "DENY",  agent, mandate, policy)           # False (verdict tamper)

An ALLOW bound under policy P returns a different guardrail_ref if recomputed against the policy_bound_ref under a rotated P'; rotation is detectable. Invalid verdicts and malformed refs are rejected, not hashed.

Conformance

conformance/spend_guardrail_lite_v1/: 2 positives (ALLOW/DENY), 4 divergence negatives (verdict / policy-rotation / agent / mandate tamper), 2 rejection negatives (invalid verdict + malformed ref), and 2 invariants. Verifier imports only stdlib + rfc8785:

pip install rfc8785
python conformance/spend_guardrail_lite_v1/verify.py

Adopters

The 4-check gate is scripts/check_v0_adoption.py (dependency + canonical hash anchor

• NOTICE + version pin → ISSUE_V0_KEY). Apply: [e-mail].

License

Apache-2.0. Copyright 2026 AlgoVoi. Preserve the NOTICE in any distribution.