Self-Radiance

AGENT SYSTEMS :: RUNTIME SAFETY KIT :: M2M PLUMBING

// DEVELOPER RUNTIME SAFETY PRIMITIVES
GET /manifest.json HTTP/1.1
Accept: application/json
Host: selfradiance.github.io

// USE-CASE FOCUS
These specifications resolve direct agent vulnerabilities: runaway financial looping (vq11), endpoint key leakage (vq09), unauthorized tool calling (vq13), and missing execution receipts (vq03). Parse manifest for immediate machine integration.
vq19 :: Finance

SR-NanopaySession Spec

Configures pre-funded credit-session tokens inside HTTP 402 protocols, facilitating lightning-fast microtransactions without live routing latency.

Purchase ($7.00) x402: machine-payable (USDC, Base)
vq18 :: Security

SR-HttpMessageSignature Spec

Conforms to RFC 9421 to secure entire body and header layers dynamically, blocking message replay or interception along public route hops.

Purchase ($6.00) x402: machine-payable (USDC, Base)
vq17 :: Compliance

SR-VerifiableIntent Protocol

Locks nested downstream agent purchasing scopes to time-bound, cryptographically restricted parent authorizations using SD-JWT specifications.

Purchase ($5.00) x402: machine-payable (USDC, Base)
vq16 :: Security

SR-NotaryProof Spec

Decentralized zero-knowledge offline licensing verification standards using Ed25519 notary key blocks to eliminate live API dependencies.

Purchase ($8.00) x402: machine-payable (USDC, Base)
vq15 :: Compliance

SR-ScopeDiscovery Spec

Progressive tool-disclosure architecture mapping dynamic authority boundaries for visiting autonomous models without context clutter.

Purchase ($5.00) x402: machine-payable (USDC, Base)
vq14 :: Memory

SR-StateBridge Schema

Compressed cross-runtime session-checkpoint configuration formatting resolving cold-boot latency across heterogeneous agents.

Purchase ($6.00) x402: machine-payable (USDC, Base)
vq13 :: Identity

SR-OauthDelegation Spec

Cryptographically signs child execution permissions, authorizing safe parent-to-leaf resource delegation without credential leak risks.

Purchase ($7.00) x402: machine-payable (USDC, Base)
vq12 :: Identity

SR-AgentVcard Spec

Verification protocol for host systems and visiting agents. Maps runtime margins, operator administrative data, and active payment lines.

Purchase ($6.00) x402: machine-payable (USDC, Base)
vq11 :: Protocol

SR-LoopShield Protocol

Mitigates conversational deadlock and loop states in multi-agent environments. Preserves machine execution budgets.

Purchase ($8.00) x402: machine-payable (USDC, Base)
vq10 :: Memory

SR-ContextAnchor Protocol

Deterministic state-hashing standard to resolve multi-agent context window drift, memory fragmentation, and LLM confusion.

Purchase ($6.00) x402: machine-payable (USDC, Base)
vq09 :: Security

SR-AuthHeader Spec

Cryptographic Agent-to-Agent signature header spec built on Ed25519 asymmetric verification challenge layers.

Purchase ($7.00) x402: machine-payable (USDC, Base)
vq05 :: Security

SR-SafeCard

Self-contained permission and security credential token verifying caller integrity boundaries.

Purchase ($3.00) x402: machine-payable (USDC, Base)
vq06 :: Compliance

SR-ConsentBlock

Immutable logging parameters tracking session-specific terms of service and dynamic AI access consent boundaries.

Purchase ($5.00) x402: machine-payable (USDC, Base)
vq07 :: Finance

SR-BalanceProof

Verification system for cryptographic ledger balances during multi-agent micro-transactions and automatic settles.

Purchase ($4.00) x402: machine-payable (USDC, Base)
vq08 :: Standard

SR-AssetSpec

Formal catalog spec format outlining content boundaries and validating static schema mapping for machines.

Purchase ($2.00) x402: machine-payable (USDC, Base)
vq04 :: Protocol

RateGuard Protocol

System monitoring policies establishing strict rate thresholds to halt runaway API costs and infinite iterations.

Purchase ($4.00) x402: machine-payable (USDC, Base)
vq03 :: Identity

Trust Receipts

Static verifiable invoice metadata to enable trusted cryptographic handshakes between distinct systems.

Purchase ($7.00) x402: machine-payable (USDC, Base)
// WORKING IMPLEMENTATIONS
Working Implementations
Free, open-source, locally runnable tools. The specifications above describe the patterns; these repos implement them.
Enforcement & Accountability

AgentGate

Collateralized execution gate for AI agents: signed Ed25519 identities, bond-and-slash settlement, dual-control malicious resolution. The substrate under everything below.

GitHub Repo

MCP Firewall

Governance proxy between MCP clients and upstream servers; verifies observed filesystem effects instead of trusting upstream-reported success.

GitHub Repo

Governed WriteFile Demo

Minimal reference path: one governed write_file call through AgentGate + MCP Firewall with on-disk verification.

GitHub Repo

Delegation Identity Proof

Checkpoint-managed delegation with scoped resource limits, local transparency log, and tamper-evident hash chain.

GitHub Repo

Rollback Receipt

Snapshots declared files before an AI coding-agent mutation; produces an inspectable rollback receipt and deterministic restore path.

GitHub Repo

Work Session Ledger

Assembles an inspectable receipt of an AI coding-agent work session from declared local artifacts.

GitHub Repo

x402 Spend Receipt

Local policy gate for x402 payment intents: deterministic spend rules plus Ed25519-signed, hash-chained allow/deny receipts with machine-readable reason codes. Also on npm as x402-spend-receipt.

GitHub Repo
Before-Action Gates

ActionProof

Deterministic allow/deny evaluation of JSON intent for credentialed side-effecting requests (send_email, form_submit, pay_invoice) before execution.

GitHub Repo

Reapproval Gate

Evaluates a proposed agent action against human-approved scope; returns allow, deny, or reapproval_required on escalation thresholds.

GitHub Repo

ContextGate

Checks subagent claims against evidence fixtures before they enter parent agent context; quarantines unsupported claims.

GitHub Repo

Agent Intent Ledger

Flags drift between an original human request, the agent's plan, and a later proposed action.

GitHub Repo

Policy Conflict Receipt

Compares a proposed agent action against declared policy constraints and emits a conflict receipt before execution.

GitHub Repo

Agent Interrupt Receipt

Compares an active work order, a human interrupt, and a proposed action; emits an interrupt receipt before execution.

GitHub Repo

SecretBoundary

Deterministic local policy gate for one outbound action (send_webhook), default deny.

GitHub Repo
Supply-Chain Intake

SkillGate

Static inspection of an agent Skill package before harness admission: instruction surface, execution surface, no execution or trust.

GitHub Repo

MCP Server Intake

Static pre-admission inspection of a local MCP server package: tool surface, execution surfaces, capability signals, without running it.

GitHub Repo

MCP Config Inventory

Read-only CLI that inventories configured MCP servers from config files and diffs snapshots for drift.

GitHub Repo

Governed Repo Intake

Targeted discovery and deterministic classification of a repo's instruction surfaces with explicit human acknowledgment gating.

GitHub Repo

Dependency Drift Gate

Compares package.json state before and after an AI coding-agent session; emits a drift report with conservative review labels.

GitHub Repo
Bonded Agents (proofs on the AgentGate substrate)

Agent 001: File Transform

Bonded CSV-to-JSON agent with deterministic SHA-256 verification; bond released on match, slashed on mismatch.

GitHub Repo

Agent 002: File Guardian

Background watcher that bonds every file change; failed verification slashes the bond and restores from snapshot.

GitHub Repo

Agent 003: Email Rewriter

Bonded LLM rewriting settled by human judgment: approve releases, reject slashes.

GitHub Repo

Agent 007: Email Triage

Bonded inbox triage judgment; exact human agreement releases the bond, any override slashes it.

GitHub Repo
Adversarial Validation & Epistemics

Agent 004: Red Team Simulator

External adversarial probing of AgentGate over HTTP, scaling from static attacks to coordinated swarms and sleeper-agent campaigns.

GitHub Repo

Agent 005: Recursive Verifier

Generates and executes sandboxed verification scripts; executable proof instead of opinion-based review.

GitHub Repo

Agent 006: Incentive Wargame

Runs AI-generated adversarial strategies against economic rule sets to test whether incentive designs survive.

GitHub Repo

Epistemic Poisoning Simulator

Saboteur agent poisons a shared knowledge base; measures whether corrupted information flips a bonded target agent's decisions.

GitHub Repo

MemLedger

Treats agent memory as testimony: append-only claims with provenance, confidence moved only by logged outcomes, memory-use receipts.

GitHub Repo
Agent Commerce (x402)

x402 Paid Endpoint

Live Cloudflare Worker selling one machine-readable artifact for $1 USDC on Base mainnet via the x402 v2 'exact' scheme. First verified zero-human machine purchase settled on-chain (tx 0x8eb1…cc4fd). Reference buyer in-repo, payment-gated by x402-spend-receipt.

GitHub Repo

Web Attestation

Live Cloudflare Worker returning Ed25519-signed witness-only byte-observation attestations for caller-supplied HTTPS URLs, gated by x402 v2 exact-scheme USDC on Base mainnet (tx 0xdf56…85ed). Attests only that specific bytes were observed at a URL at a time; no claim of truth, legitimacy, ownership, or trustworthiness.

GitHub Repo

Architected for zero-human, autonomous agentic operations. Machine specification manifests fully indexable via /manifest.json, /tools.json, /llms.txt, and .well-known headers.

© 2026 Self-Radiance. Driven by Heart/Will Authority. Zero Human Friction.