agent-safe-spl 0.3.0

SPL (Safe Policy Lisp) evaluator for Agent-Safe capability tokens. 150 lines, zero deps, microseconds.

Agent-Safe SPL — Python SDK

Python implementation of the SPL (Safe Policy Lisp) evaluator for Agent-Safe capability tokens.

Install

pip install -e .               # core (zero runtime deps)
pip install -e ".[crypto]"     # with Ed25519/Merkle/hash-chain support
pip install -e ".[dev]"        # with test dependencies

Usage

As a library

from spl import parse, verify

policy = parse('(and (= (get req "action") "read") (<= (get req "amount") 100))')
request = {"action": "read", "amount": 50}
env = {
    "vars": {},
    "per_day_count": lambda action, day: 0,
    "crypto": {
        "dpop_ok": lambda: True,
        "merkle_ok": lambda t: True,
        "vrf_ok": lambda d, a: True,
        "thresh_ok": lambda: True,
    },
}

result = verify(policy, request, env)
print("ALLOW" if result["allow"] else "DENY")

As a CLI

python -m spl examples/policies/family_gifts.spl examples/requests/gift_50_niece.json
# → ALLOW

Tests

pip install -e ".[dev]"
pytest tests/ -v

Requirements

• Python 3.10+
• Zero runtime dependencies for core evaluator
• cryptography package optional for real Ed25519 verification