Kinetic Gain Protocol Suite

A family of twelve open JSON specifications for the answer-engine and agent era — plus a fifteen-repo implementation stack that consumes them. Five core specs · three EdTech extensions · one HealthTech extension · one cross-cutting incident-disclosure spec · one buyer-side procurement Decision Card (v0.3 — adds data_vault_targets + retention_envelope) · one InsurTech Claims Decision Card · one unified visualizer · one operator console · one unified MCP server · 107 live properties on kineticgain.com · all AGPL-3.0 (specs) + MIT (implementations).

Public front door: suite.kineticgain.com.

Kinetic Gain Operator Console — mission-control for the whole Suite: an interactive topology mesh (v0.2 — runtime-gate overlays for the MCP Permission Broker / Azure Governance Bridge / SQL Contract Enforcer, ed25519 signature posture, and blast-radius tracing across the hash-chained spine), a configurable SRE operator dashboard (savable layouts, simulation throttle, per-channel latency sliders, PDF export), and live audit-stream visualization. (repo)

We publish the specs AND we score against them. Live receipt at kineticgain.com/.well-known/pulse-receipt.json, refreshed weekly by procurement-pulse-action on kineticgain.com's own GitHub Actions.

This repository is the single landing for the Suite — both for the specs (the "what to publish" layer) and for the implementation stack that consumes them (the "what to actually run" layer). Each spec lives in its own repo with full normative text, JSON Schema, examples, and a permissive cross-link table. This meta-repo is where you start when you want the map.

🧭 The suite at a glance

┌──────────────────── Core (Answer Engine / Agent layer) ────────────────────┐
│                                                                            │
│  AEO Protocol ──── entity declaration at /.well-known/aeo.json             │
│  Prompt Provenance ─ versioned, lineaged, reviewable LLM prompts          │
│  Agent Cards ───── capability + refusal disclosure for AI agents          │
│  AI Evidence Format ─ structured citations for LLM-generated claims       │
│  MCP Tool Cards ── per-tool disclosure for MCP servers                    │
│                                                                            │
└────────────────────────────────────────────────────────────────────────────┘
                                    │
                                    │ pairs with
                                    ▼
┌──────────────────── EdTech trio (vendor / district / student) ─────────────┐
│                                                                            │
│  AI Tutor Cards ── what an AI tutor does (vendor-side)                    │
│      ↕                                                                     │
│  Classroom AI AUP ─ what AI use is permitted (district / school / course) │
│      ↕                                                                     │
│  Student AI Disclosure ─ what the student actually did (per artifact)     │
│                                                                            │
└────────────────────────────────────────────────────────────────────────────┘

┌──────────────── HealthTech extension (vendor-side disclosure) ─────────────┐
│                                                                            │
│  Clinical AI Disclosure ── what a healthcare AI system does               │
│  HIPAA / FDA / SaMD posture · bias audits · EHR (FHIR / CDS Hooks)        │
│  references → Agent Card · AI Evidence · AI Incident Card                  │
│                                                                            │
└────────────────────────────────────────────────────────────────────────────┘
                                    │
                                    │ when things break
                                    ▼
┌──────────── Cross-cutting (vendor-published, references all) ──────────────┐
│                                                                            │
│  AI Incident Card ── post-incident disclosure ("CVE for AI agents")       │
│  references → Agent / Tutor / Tool Cards · Prompt Provenance · AI Evidence │
│                                                                            │
└────────────────────────────────────────────────────────────────────────────┘
                                    │
                                    │ buyer-side (when approving / rejecting)
                                    ▼
┌──────────── Cross-cutting (buyer-published, references vendor docs) ───────┐
│                                                                            │
│  AI Procurement Decision Card ── buyer's signed-off position on a vendor  │
│  status: approved / approved-with-conditions / rejected / withdrawn       │
│  references → AEO · Agent / Tool / Tutor / Clinical Cards · AI Evidence    │
│                                                                            │
└────────────────────────────────────────────────────────────────────────────┘

The Suite is bilateral: vendors publish disclosures (the first six rows); buyers publish Decision Cards that approve or reject those disclosures. AI Incident Cards close the loop when something breaks.

📐 Specifications

Every spec carries a top-level <name>_version field. The unified visualizer detects which spec a document is in by inspecting that single field.

Spec	Repo	Detect via	Well-known path
AEO Protocol	`aeo-protocol-spec`	`aeo_version`	`/.well-known/aeo.json`
Prompt Provenance	`prompt-provenance-spec`	`provenance_version`	—
Agent Cards	`agent-cards-spec`	`agent_card_version`	`/.well-known/agents/<agent_id>.json`
AI Evidence Format	`ai-evidence-format-spec`	`evidence_version`	—
MCP Tool Cards	`mcp-tool-card-spec`	`tool_card_version`	`/.well-known/mcp-tools/<tool_name>.json`
AI Tutor Cards (EdTech)	`ai-tutor-card-spec`	`tutor_card_version`	`/.well-known/tutors/<tutor_id>.json`
Student AI Disclosure (EdTech)	`student-ai-disclosure-spec`	`disclosure_version`	— (travels with the artifact)
Classroom AI AUP (EdTech)	`classroom-ai-aup-spec`	`aup_version`	`/.well-known/ai-aup.json`
Clinical AI Disclosure (HealthTech)	`clinical-ai-disclosure-spec`	`clinical_ai_card_version`	`/.well-known/clinical-ai/<system_id>.json`
AI Incident Card (cross-cutting, vendor-side)	`ai-incident-card-spec`	`incident_card_version`	`/.well-known/ai-incidents/<id>.json` (+ index at `/.well-known/ai-incidents.json`)
AI Procurement Decision Card (cross-cutting, buyer-side)	`ai-procurement-decision-spec`	`decision_card_version`	`/.well-known/procurement-decisions/<id>.json`
AI Claims Decision Card (InsurTech, buyer-side)	`ai-claims-decision-card-spec`	`claims_card_version`	(travels with the artifact)

All twelve: AGPL-3.0 spec text, freely implementable, v0.1 draft, JSON Schema draft 2020-12, tagged kinetic-gain-protocol-suite. The NIST AI RMF crosswalk maps every spec — plus the implementation tooling below — to specific NIST AI RMF subcategories.

🧱 Reference implementations

The Suite now has real implementation tooling attached to several of its disclosure and distribution patterns:

MCP Tool Cards
- mcp-tool-card-generator — emits Tool Card disclosures directly from MCP tools/list output
- mcp-registry-risk-scanner — validates manifest and registry posture around published MCP servers
- mcp-tools-diff — detects tools/list drift and breaking-change posture over time
Agent Cards
- agent-card-runtime-adapters — maps Agent Cards into runtime adapters for OpenAI, Anthropic, and Vercel execution layers
AI Evidence Format / Evidence Ledger
- rag-evidence-graph — corpus-level citation graph and evidence coverage analysis
- rag-evidence-trace-linker — per-call trace linkage for evidence integrity on RAG and generation flows
/.well-known/ distribution pattern
- wellknown-index-aggregator — builds and validates /.well-known/index.json for multi-doc publisher estates
- governance-disclosure-operator — Kubernetes-native publisher for governance disclosures

🛠️ Suite × Implementations

The Suite is a set of specs. This section is the software that consumes them — fifteen repos across Tiers A–E, all CI-green, semver-tagged, MIT-licensed, with five cross-ecosystem hooks tying them together. Grouped by the buyer most likely to land on the repo first.

🕸️ How it composes

flowchart TB
    classDef spec fill:#10b981,stroke:#065f46,color:#fff,stroke-width:2px
    classDef hook fill:#3b82f6,stroke:#1e40af,color:#fff,stroke-width:2px
    classDef sup fill:#f3f4f6,stroke:#6b7280,color:#1f2937
    classDef stream fill:#f59e0b,stroke:#92400e,color:#fff
    classDef mcp fill:#a855f7,stroke:#581c87,color:#fff,stroke-width:2px

    SPECS["📐 11 Suite specs<br/>AEO · Agent · Tool · Tutor · AUP · Disclosure<br/>Evidence · Provenance · Clinical · Incident · Decision"]:::spec

    SPECS -->|"#1 ingest Suite docs"| PDA["procurement-decision-api<br/>drafts Decision Cards"]:::hook
    PDA -->|"#2 conditions → runtime gates"| PAC["policy-as-code-engine<br/>PolicyBundle enforcement"]:::hook
    PDA -->|"#3 extract owners"| DCR["data-contract-registry<br/>schema + SLAs"]:::hook
    DCR -->|"#4 streaming CSV check"| CDQ["csv-data-quality-rs<br/>row-by-row validation"]:::hook
    DCR -->|"#5 contract → DDL"| SCE["sql-contract-enforcer<br/>cross-dialect table constraints"]:::hook

    SPECS -.->|sign + verify| HA["hash-attestation-rs<br/>ed25519 over canonical hash"]:::sup
    SPECS -.->|drift detection| AVS["aeo-validator-service<br/>always-on validation"]:::sup
    AVS -.->|JSONL feed| AGE["aeo-graph-explorer-rs<br/>graph-query layer #5"]:::sup
    SPECS -.->|incident → plan| ICR["incident-correlation-rs<br/>Suite-graph BFS"]:::sup
    ICR -.->|drives| PAC

    PDA --> AS
    PAC --> AS
    DCR --> AS
    AVS --> AS
    ICR --> AS
    HA --> AS
    AS["📋 audit-stream-py<br/>hash-chained tamper-evident spine"]:::stream

    SPECS ==>|spec tools| MCP
    PDA ==>|preview tools| MCP
    AS ==>|event tools| MCP
    HA ==>|verify tools| MCP
    MCP["🤖 mcp-kinetic-gain v0.9.1<br/>75 tools · one Claude Desktop config entry"]:::mcp

Green = the spec foundation. Blue = the five cross-ecosystem hooks that make this a stack rather than a pile of repos. Grey = supporting tools that feed either side. Amber = the tamper-evident audit-stream spine every governance moment writes to. Purple = the unified MCP surface that exposes the whole thing to Claude.

📋 The audit-stream spine — eleven producers, five runtimes

Zoom in on the amber spine: every governance moment in the stack writes to one hash-chained, tamper-evident log via audit-stream-py. Same opt-in env-var contract (AUDIT_STREAM_URL) across every producer; same best-effort semantics (a failed POST is logged, never raised). The original seven (four FastAPI services + three Rust crates) are now joined by four runtime + data-tier producers — mcp-permission-broker and azure-openai-governance-bridge (tool-invocation gates), pg-audit-stream-extension (Postgres CRUD via pg_notify), and wp-kinetic-gain-audit (WordPress/MySQL) — so the spine now spans Python, Rust, PL/pgSQL, PHP, and Azure Functions. One verifiable narrative an auditor can replay end-to-end, no matter which tier emitted the event.

flowchart LR
    classDef pyprod fill:#3b82f6,stroke:#1e40af,color:#fff,stroke-width:2px
    classDef rsprod fill:#dea584,stroke:#92400e,color:#1f2937,stroke-width:2px
    classDef spine fill:#f59e0b,stroke:#92400e,color:#fff,stroke-width:3px
    classDef sink fill:#f3f4f6,stroke:#6b7280,color:#1f2937

    PDA["procurement-decision-api<br/>Python · FastAPI"]:::pyprod
    AVS["aeo-validator-service<br/>Python · FastAPI"]:::pyprod
    PCE["policy-as-code-engine<br/>Python · FastAPI"]:::pyprod
    DCR["data-contract-registry<br/>Python · FastAPI"]:::pyprod
    HA["hash-attestation<br/>Rust · crypto library"]:::rsprod
    ICR["incident-correlation<br/>Rust · graph library"]:::rsprod
    AGE["aeo-graph-explorer<br/>Rust · axum service"]:::rsprod
    MPB["mcp-permission-broker +<br/>azure-openai-governance-bridge<br/>Python · runtime gates"]:::pyprod
    PGX["pg-audit-stream-extension<br/>PL/pgSQL · pg_notify"]:::rsprod
    WPA["wp-kinetic-gain-audit<br/>PHP · WordPress/MySQL"]:::rsprod

    PDA -->|"decision_card_drafted"| AS
    AVS -->|"watch_created<br/>watch_drifted<br/>watch_validity_flipped"| AS
    PCE -->|"policy_bundle_registered<br/>request_allowed<br/>request_denied"| AS
    DCR -->|"contract_promoted<br/>contract_deprecated<br/>contract_compatibility_failed"| AS
    HA -->|"attestation_signed<br/>attestation_verified<br/>attestation_failed"| AS
    ICR -->|"incident_correlated<br/>incident_correlation_failed"| AS
    AGE -->|"graph_ingested<br/>graph_ingest_failed"| AS
    MPB -->|"tool_invocation_allowed<br/>tool_invocation_denied<br/>tool_invocation_required_approval"| AS
    PGX -->|"&lt;configured kind&gt; on table CRUD"| AS
    WPA -->|"content_published<br/>plugin_activated<br/>user_role_changed"| AS

    AS{{"📋 audit-stream-py<br/>hash-chained · tamper-evident<br/>SSE live tail · REST query · GET /verify"}}:::spine

    AS -->|GET /events/stream| LT["governance dashboards<br/>(live tail)"]:::sink
    AS -->|GET /events| Q["compliance evidence<br/>(REST query)"]:::sink
    AS -->|GET /verify| V["auditor replay<br/>(walk the chain)"]:::sink

Blue = Python FastAPI producers. Tan = Rust producers (two libraries gated behind --features audit-stream so library consumers can strip out the HTTP dep, one axum service with the feature on by default). Amber = the spine itself. Grey = the three downstream surfaces auditors and operators consume.

Adding the next producer is a ~60-line module: copy the audit_stream shape (Python, Rust, PL/pgSQL, or PHP), pick your event kinds, point at AUDIT_STREAM_URL. The data-tier producers prove the point — pg-audit-stream-extension catches direct DML the application path would miss, and wp-kinetic-gain-audit brings the same tamper-evident chain to any WordPress estate. Next natural candidates: slo-budget-tracker (slo_burn_started / slo_recovered) or reliability-toolkit-rs (breaker_opened / breaker_recovered).

🛒 Procurement reviewer / buyer-side governance

Repo	Lang	What it does
`procurement-decision-api`	Python · FastAPI	Drafts AI Procurement Decision Cards from a buyer rubric and a set of vendor Suite documents (AEO + agent-card + tool-card + ai-evidence + …). The first cross-ecosystem bridge in the portfolio — Suite × Decision Intelligence.
`policy-as-code-engine`	Python · FastAPI	Declarative policy evaluator. Headline: `POST /bundles/from-decision-card` turns a Decision Card's `conditions[]` into a runtime-enforceable PolicyBundle. Closes the loop from "buyer signed off" to "request gated." Cross-ecosystem hook #2.
`hash-attestation-rs`	Rust · ed25519	Sign and verify Suite documents with ed25519 over the canonical-hash convention every other Suite repo already uses. The missing "this AEO actually came from the vendor" layer.

🌐 AEO consumer / spec implementer

Repo	Lang	What it does
`aeo-validator-service`	Python · FastAPI	Always-on HTTP validator for AEO + all 11 Suite docs. Auto-detects the spec via `_version` sniffing, hashes canonically, tracks drift* across re-checks (`POST /watches/{id}/recheck` returns a structured `DriftReport`). Layer 4 of the AEO Reference Stack.
`aeo-graph-explorer-rs`	Rust · axum · petgraph	HTTP graph-query service over `aeo-crawler` JSONL output. `GET /neighbors`, `GET /shortest-path`, `GET /find-by-claim`, atomic `POST /ingest`. Layer 5 of the AEO Reference Stack.
`incident-correlation-rs`	Rust · petgraph	Walks the Suite graph from an `IncidentCard` and emits a structured remediation plan: `DecisionCard → RecheckPolicy`, `Vendor → RequestReview`, AEO/agent/tool → `Revalidate`.

🗃️ Data team

Repo	Lang	What it does
`data-contract-registry`	Python · FastAPI	Schema registry with semver versioning, compatibility checks (backward / forward / full), declared owners, freshness SLAs. `POST /contracts/owners/from-decision-card` pulls Owner records out of a Procurement Decision Card. Cross-ecosystem hook #3.
`csv-data-quality-rs`	Rust · tokio · csv	Streaming CSV validator against a `data-contract-registry` contract. Async, row-by-row, structured violation report (`required` / `bad_type` / `enum_mismatch` / `column_count_mismatch` / `invalid_json`). Cross-ecosystem hook #4.
`sql-contract-enforcer`	Python · SQL	Turns a `data-contract-registry` contract into enforceable cross-dialect DDL (CHECK / NOT NULL / UNIQUE / PK / FK) for Postgres, MySQL, Snowflake, BigQuery, plus a contract-vs-schema checker for CI. Dialect-aware (BigQuery demotes CHECK/UNIQUE to comments + PK/FK to NOT ENFORCED). Cross-ecosystem hook #5.

🛡️ SRE / Platform reliability stack

Ten repos that compose into a single layered reliability story: identity → rate limits → canary → registry → SLO budget → Rust primitives → feature flags → shadow traffic → tamper-evident audit log.

Repo	Lang	What it does
`slo-budget-tracker`	Python · FastAPI · Prometheus	SLO + error-budget library, multi-window burn-rate alerts (SRE workbook threshold 14.4).
`reliability-toolkit-rs`	Rust · Tokio	Token-bucket rate limiter · 3-state circuit breaker · exponential-backoff retry with jitter · bulkhead.
`feature-flag-rs`	Rust · Tokio	Server-side flag eval — targeting rules, sticky percentage rollouts (SHA-256 bucketing, no RNG), hot reload.
`request-shadow-rs`	Rust · Tokio	Async request mirroring with sampling + divergence detection. The SRE primitive for safe migrations.
`audit-stream-py`	Python · FastAPI · SSE	Append-only governance event stream, hash-chained for tamper-evidence. Every portfolio repo can produce events here.
(Plus 5 earlier reliability repos)	Python	`rate-limit-shield` · `identity-mesh` · `agent-canary` · `model-registry-pro` — defense-in-depth predecessors.

🤖 MCP / Claude integrator

Repo	Lang	What it does
`mcp-kinetic-gain`	TypeScript	Unified Suite MCP server — 75 tools across 12 specs + DefenseTech 6-pack (v0.9.1, on the official MCP Registry). One Claude Desktop config entry. Headline tools: `aup_check_compliance` (joins AUP + Disclosure into one allow/deny call); `decision_card_validate` (enforces the full Decision Card conditional rule set); `defensetech_vault_resolve_3axis` (resolves CUI × export-control × foreign-person to the most-restrictive vault policy).
`mcp-reliability-toolkit`	TypeScript	Reliability MCP server — `compute_slo_burn`, `design_rate_limiter`, `design_circuit_breaker`, `compose_reliability_pattern`. Same math as `slo-budget-tracker`; emits Python + Rust configs.
`mcp-decision-intelligence`	TypeScript	Decision Intelligence MCP server — `validate_decision_card`, `preview_policy_bundle`, `plan_incident_remediation`, `check_contract_compatibility`. Read-only preview of what the live Python/Rust services would compute.

🚦 Runtime enforcement — turning "buyer signed off" into "request denied"

The Decision Card layer decides; these gates enforce, at the moment a tool is invoked, and write every verdict to the spine:

Repo	Lang	What it does
`mcp-permission-broker`	Python	Runtime gate between a Decision Card and an MCP tool call. Composes Decision Card conditions into deny-trumps-allow PolicyBundles; emits `tool_invocation_*` to the spine.
`azure-openai-governance-bridge`	Python · Azure Functions · Bicep	The Azure-native sibling — an Azure Function in front of Azure OpenAI enforcing the same PolicyBundle contract on every chat-completion call (deployment + each declared tool). Puts the Suite's governance on the data path enterprises actually run AI on.

The five cross-ecosystem hooks

What makes the stack a stack rather than a list of repos:

procurement-decision-api → Suite documents. Ingests AEO / agent-card / tool-card / ai-evidence by URL; emits a Decision Card. (Suite × Decision Intelligence.)
policy-as-code-engine → procurement-decision-api. POST /bundles/from-decision-card turns approve / reject / approve-with-conditions into runtime-enforceable allow / deny / per-condition gates.
data-contract-registry → procurement-decision-api. POST /contracts/owners/from-decision-card extracts buyer + decision_maker into Owner records so freshly registered contracts carry paging info nobody re-types.
csv-data-quality-rs → data-contract-registry. Streaming CSV validator against a registered contract — producers prove their output matches, row by row.
sql-contract-enforcer → data-contract-registry. Compiles the same contract into cross-dialect DDL (CHECK / NOT NULL / UNIQUE / PK / FK across Postgres, MySQL, Snowflake, BigQuery) — enforcing at the table boundary what hook #4 validates row-wise.

The implementation stack is independently usable — any repo composes with the rest, none requires it — and the Decision Card now enforces at three layers: the MCP tool call (mcp-permission-broker), the Azure OpenAI call (azure-openai-governance-bridge), and the database table (sql-contract-enforcer).

🔌 One MCP server. 75 tools. Twelve specs + DefenseTech.

mcp-kinetic-gain (v0.9.1, on the official MCP Registry) is the unified Model Context Protocol server exposing every runtime Kinetic Gain spec as callable tools. One Claude Desktop / Cursor / MCP-client config entry; 75 tools across all twelve specs plus the DefenseTech 6-pack (3-axis vault resolver + CUI/ITAR/DFARS invariants + CMMC evidence summarizer + Incident Card event classifier).

Spec	Tools
AEO Protocol	`aeo_fetch` · `aeo_inspect` · `aeo_get_claim` · `aeo_well_known_url`
Prompt Provenance	`prompt_provenance_validate` · `prompt_provenance_inspect` · `prompt_provenance_eval_result`
Agent Cards	`agent_card_well_known_url` · `agent_card_inspect` · `agent_card_tool_disclosure` · `agent_card_validate`
AI Evidence Format	`ai_evidence_validate` · `ai_evidence_inspect` · `ai_evidence_verify_hash`
MCP Tool Cards	`tool_card_well_known_url` · `tool_card_inspect` · `tool_card_tested_with` · `tool_card_validate`
AI Tutor Cards	`tutor_card_well_known_url` · `tutor_card_fetch` · `tutor_card_validate` · `tutor_card_inspect` · `tutor_card_subject_check` · `tutor_card_coppa_check`
Student AI Disclosure	`disclosure_validate` · `disclosure_inspect` · `disclosure_verify_artifact_hash` · `disclosure_verify_prompt_hash` · `disclosure_aup_check`
Classroom AI AUP	`aup_well_known_url` · `aup_fetch` · `aup_validate` · `aup_inspect` · `aup_check_compliance`
Clinical AI Disclosure	`clinical_ai_well_known_url` · `clinical_ai_validate` · `clinical_ai_inspect` · `clinical_ai_samd_check`
AI Incident Card	`incident_validate` · `incident_inspect` · `incident_index_fetch` · `incident_affected_walk`
AI Procurement Decision Card	`decision_card_validate` · `decision_card_inspect` · `decision_card_conditions` · `decision_card_signature_check`

172 tests pass · typecheck clean · stdio MCP server, drops into any MCP-compatible client. Sibling MCP servers for the Reliability Stack (mcp-reliability-toolkit) and Decision Intelligence (mcp-decision-intelligence) exist for narrower use cases.

🖼️ One visualizer. Eleven specs.

kinetic-gain-visualizer auto-detects the spec from the top-level *_version field and renders the appropriate procurement-grade view. Live on GitHub Pages.

Visualize — the auto-detected renderer (FERPA / COPPA / GDPR badges for tutor cards, role-tone pills + artifact-hash binding for disclosures, vendor-requirements card in dark/authority tone for AUPs, full conditional-rule preview for Decision Cards, etc.)
Editor — paste any spec document and watch the right view light up
Architecture — the 12-spec map
Tools — searchable catalog of all 63 MCP tools
About — detection model + cross-links

⚙️ Suite governance Actions — 9-action CI ecosystem

A pre-built set of GitHub Actions for PR-gating governance documents before merge. Drop them into any repo that holds AgentCards / Tool Cards / prompt-provenance / evidence bundles / OTel spans and they post a Markdown summary as a PR comment + fail the build on high-severity findings.

Per-protocol fleet-summary gates (4)

Action	Wraps	Catches
`agent-card-fleet-summary-action`	`agent-card-fleet-summary`	autonomous-without-IRU, destructive-on-non-autonomous, persistent-memory-without-refusal-taxonomy
`mcp-tool-card-fleet-summary-action`	`mcp-tool-card-fleet-summary`	destructive-without-human-approval, high-PII-without-rate-limit, writes-secrets-without-audit
`prompt-provenance-fleet-summary-action`	`prompt-provenance-fleet-summary`	approved-without-evaluations, no-reviewers, failing-eval-on-approved, deprecated-still-referenced
`evidence-bundle-fleet-summary-action`	`evidence-bundle-fleet-summary`	bundle-expired, unsigned-regulated-bundle, cross-bundle-hash-collision, oversized-bundle

Cross-protocol governance gates (3)

Action	Catches
`kg-suite-spec-version-tracker-action`	Version drift — same protocol on two different spec versions in one fleet
`kg-suite-conformance-runner-action`	Structural conformance — missing required top-level blocks per spec
`kg-suite-canonicalize-action`	Silent edits — doc's canonical sha256 changed without a version bump

Cost + cluster gates (2 — pre-existing)

Action	Catches
`llm-cost-rollup-action`	GenAI FinOps — budget breaches across OTel cost-annotated spans
`k8s-pre-merge-action`	K8s pre-merge — deprecated APIs, RBAC over-scope, Pod Security, Helm values coverage

All nine are composite Node 20 actions with dist/ committed for SHA/tag pinning — no build step needed in the consuming workflow.

🛡️ Testing artifact

Repo What it does

prompt-injection-bench Open 30-attack prompt-injection corpus + Python harness. Every record carries an agent_card_refusal_categories back-ref to the Agent Card refusal_taxonomy[].category it tests. A vendor can grep their declared categories against the corpus to verify their stated commitments hold under attack — and failed runs become natural inputs for AI Incident Cards. Visual harness live at bench.kineticgain.com.

The bench is not a twelfth spec — it's the testing-counterpart to the disclosure layer. The Suite tells you what an agent should refuse; the bench tells you whether it actually does.

🌐 Live properties — 32 total

Hubs + tools (8)

URL	What it serves
suite.kineticgain.com	Canonical front door for the entire Suite — 12-spec map, full spec table, two-front-doors section, NIST RMF crosswalk
docs.kineticgain.com	Quickstart hub — per-role guides + canonical `/.well-known/` path map
directory.kineticgain.com	Vendor directory — curated list of domains publishing Kinetic Gain documents
examples.kineticgain.com	Examples gallery — sidebar of 11 specs, click for canonical example with JSON highlight
walker.kineticgain.com	well-known-walker — paste any domain, see every Kinetic Gain disclosure it publishes
bench.kineticgain.com	prompt-injection-bench — paste a JSONL transcript, see pass rates by category and severity
pulse.kineticgain.com	AI Procurement Pulse — quarterly research index of vendor disclosure across the open internet. Issue #1 "The Zero Baseline" is live (powered by `procurement-pulse-engine` + `well-known-probe-js`)
console.kineticgain.com	Operator Console — mission-control for the Suite: interactive topology mesh (v0.2 — runtime-gate overlays, ed25519 signature posture, blast-radius tracing), configurable SRE operator dashboard, live audit-stream visualization, PDF export (repo)

Per-spec landings (11)

URL	Spec
aeo.kineticgain.com	AEO Protocol — interactive visualizer
prompts.kineticgain.com	Prompt Provenance
agents.kineticgain.com	Agent Cards
evidence.kineticgain.com	AI Evidence Format
toolcards.kineticgain.com	MCP Tool Cards
tutor.kineticgain.com	AI Tutor Cards (EdTech)
student.kineticgain.com	Student AI Disclosure (EdTech)
aup.kineticgain.com	Classroom AI AUP (EdTech)
clinical.kineticgain.com	Clinical AI Disclosure (HealthTech)
incidents.kineticgain.com	AI Incident Card (cross-cutting, vendor-side)
decisions.kineticgain.com	AI Procurement Decision Card (cross-cutting, buyer-side)

Earlier product surfaces (5)

URL	What it does
gv.kineticgain.com	GitVisualizer — visual portfolio intelligence for any GitHub user
mcp.kineticgain.com	MCP Sentinel — governance dashboard for MCP servers
rag.kineticgain.com	RAG Sentinel — hallucination, drift, citation quality monitoring
observe.kineticgain.com	AgentObserve — operator console for AI agent fleets
mizcausevic-dev.github.io/kinetic-gain-visualizer	Unified visualizer (GitHub Pages, all 11 specs)

Cloud Identity, Platform, FinOps & Threat Detection Governance lane (8)

Production-hardened (v1.0-prod) synthetic-data operator consoles covering the multi-cloud admin stack — Microsoft, AWS, GCP, Azure. AGPL-3.0-or-later, dual-Node CI, dependabot, 95%+ statement coverage, each deployed on its own GitHub Pages subdomain.

URL	What it does	Repo
entra.kineticgain.com	Microsoft Entra access reviews, privileged-role auto-approval drift, reviewer self-review detection, decision-overdue posture	`entra-access-review-control-plane`
intune.kineticgain.com	Microsoft Intune device compliance, jailbreak/root detection, encryption gaps, OS-drift, stale check-ins, BYOD scope	`intune-device-compliance-ops`
retention.kineticgain.com	Microsoft 365 Purview retention coverage and eDiscovery custodian / hold orchestration	`m365-retention-case-orchestrator`
aws.kineticgain.com	AWS IAM Access Analyzer posture, public-access bindings, cross-account trust, remediation sequencing	`aws-iam-access-analyzer-console`
guardduty.kineticgain.com	AWS GuardDuty detector posture, threat-finding triage, credential exfiltration / crypto-mining / anomalous-API behavior, response sequencing	`aws-guardduty-triage-board`
gcp.kineticgain.com	GCP IAM snapshot drift, public `allUsers` bindings, `roles/editor` creep, service-account token-creator grants, org-policy mismatch	`gcp-iam-policy-diff-lab`
billing.kineticgain.com	GCP billing-anomaly routing, budget breaches, spend-spike escalation, idle commitments, unlabeled-cost drift, billing-export gaps	`gcp-billing-anomaly-router`
zone.kineticgain.com	Azure landing-zone baseline drift, owner-role drift, missing deny assignments, disabled Defender, diagnostics gaps, route bypass	`azure-landing-zone-drift-radar`

🪜 Quickstart paths

Pick the entry that matches what you want to do:

I want to declare my entity / agent / tool. Start at the relevant spec's examples/ folder, fork an example, validate against the schema, serve at the well-known URL.

I want to consume Kinetic Gain documents from an LLM agent. Install mcp-kinetic-gain, add one entry to your Claude Desktop config, get 75 tools.

I want to see what a document looks like. Open the unified visualizer, pick an example from the Editor view.

I'm a school district choosing AI vendors. Read the Classroom AI AUP spec. Author your AUP. Require Tutor Cards from vendors. Require Student AI Disclosures from learners. Three JSON documents, two joins, one allow/deny answer per submission.

I'm a vendor trying to sell into K-12. Publish a Tutor Card at /.well-known/tutors/<id>.json. A district AUP can then validate your card against its vendor_requirements in milliseconds.

I'm a healthcare AI vendor selling into hospitals. Publish a Clinical AI Disclosure at /.well-known/clinical-ai/<system_id>.json with your FDA / SaMD / HIPAA / EHR-integration posture. A CMIO can read it in seconds; the bias-audit URI is procurement-blocking for SaMD class II+.

I'm a procurement reviewer evaluating an AI vendor. Stand up procurement-decision-api, feed it the vendor's Suite docs + your rubric, get back a Draft Decision Card. Sign it. Want runtime enforcement? Send the signed card to policy-as-code-engine's POST /bundles/from-decision-card — every condition becomes a deny-by-default gate.

I'm an SRE adopting the reliability stack. slo-budget-tracker for the SLO math, reliability-toolkit-rs for rate-limiter + breaker + retry + bulkhead, request-shadow-rs for migrations, audit-stream-py as the tamper-evident spine.

I'm a data team enforcing contracts. data-contract-registry to manage contracts (semver + compatibility checks). csv-data-quality-rs in CI: load the contract, validate the produced CSV, fail the build on violations.

An agent of mine misbehaved and I need to disclose. Publish an AI Incident Card at /.well-known/ai-incidents/<id>.json and add it to your /.well-known/ai-incidents.json index. Run incident-correlation-rs to walk the Suite graph and produce a remediation plan in one pass.

I want to verify my agent's declared refusals hold under attack. Run prompt-injection-bench against your agent. Every attack carries an agent_card_refusal_categories back-ref to your Agent Card's refusal_taxonomy[].category values. Drop the result into your Agent Card's evaluations[] field.

📦 AEO Reference Stack (the depth example, 5 layers)

The AEO Protocol is the oldest spec in the suite and has the most complete tooling. It's the canonical depth example — every layer you'd want for any spec.

Layer	Repos
1. SDKs	`aeo-sdk-python` (live on PyPI) · `aeo-sdk-typescript` · `aeo-sdk-rust` · `aeo-sdk-go` · `aeo-sdk-swift`
2. CLI	`aeo-cli` — `aeo validate / fetch / inspect / claim` against a live well-known URL
3. Crawler	`aeo-crawler` — BFS over AEO graphs, JSONL output
4. Validator service	`aeo-validator-service` — always-on HTTP validator for all 11 Suite docs; tracks drift via `POST /watches/{id}/recheck`
5. Graph explorer	`aeo-graph-explorer-rs` — Rust + axum + petgraph graph-query service over the crawler's JSONL output
MCP server (AEO-only)	`mcp-aeo-server` — superseded by `mcp-kinetic-gain` but useful as a single-spec install

The same five-layer pattern is now feasible for other specs: the Tutor Card and Decision Card both have clear paths to crawler + validator + explorer implementations once vendor adoption justifies the build.

🤝 Status & contribution

v0.1 draft across all twelve specs · v0.1.0 across all fifteen implementation repos. Stable enough to publish; designed to evolve. Issues and pull requests welcome on any repo. Discussion of cross-spec concerns happens here in this meta-repo.

A future v0.2 sweep will probably add: detached cryptographic signing across the family (foundation already shipped in hash-attestation-rs), an automated kinetic-gain-protocol-suite validator that walks document references end-to-end, and a conformance suite for MCP servers claiming Kinetic Gain support.

📜 License

Specifications, JSON Schemas, examples, and this meta-README: AGPL-3.0. Implementation repos (procurement-decision-api, policy-as-code-engine, slo-budget-tracker, reliability-toolkit-rs, audit-stream-py, hash-attestation-rs, and the rest): MIT. Supporting sites are unrestricted under their own licenses.

👤 Author

Miz Causevic · Enterprise Technologist · Boston, MA GitHub · LinkedIn · Kinetic Gain · Medium · Skills

Name		Name	Last commit message	Last commit date
Latest commit History 33 Commits
.github		.github
docs		docs
CHANGELOG.md		CHANGELOG.md
CROSS_MODEL_BRIEF.md		CROSS_MODEL_BRIEF.md
LICENSE		LICENSE
README.md		README.md

Navigation Menu

Repository files navigation

Kinetic Gain Protocol Suite

🧭 The suite at a glance

📐 Specifications

🧱 Reference implementations

🛠️ Suite × Implementations

🕸️ How it composes

📋 The audit-stream spine — eleven producers, five runtimes

🛒 Procurement reviewer / buyer-side governance

🌐 AEO consumer / spec implementer

🗃️ Data team

🛡️ SRE / Platform reliability stack

🤖 MCP / Claude integrator

🚦 Runtime enforcement — turning "buyer signed off" into "request denied"

The five cross-ecosystem hooks

🔌 One MCP server. 75 tools. Twelve specs + DefenseTech.

🖼️ One visualizer. Eleven specs.

⚙️ Suite governance Actions — 9-action CI ecosystem

Per-protocol fleet-summary gates (4)

Cross-protocol governance gates (3)

Cost + cluster gates (2 — pre-existing)

🛡️ Testing artifact

🌐 Live properties — 32 total

Hubs + tools (8)

Per-spec landings (11)

Earlier product surfaces (5)

Cloud Identity, Platform, FinOps & Threat Detection Governance lane (8)

🪜 Quickstart paths

📦 AEO Reference Stack (the depth example, 5 layers)

🤝 Status & contribution

📜 License

👤 Author

About

Releases

Packages

Used by

Contributors

Folders and files

Latest commit

History

Repository files navigation

Kinetic Gain Protocol Suite

🧭 The suite at a glance

📐 Specifications

🧱 Reference implementations

🛠️ Suite × Implementations

🕸️ How it composes

📋 The audit-stream spine — eleven producers, five runtimes

🛒 Procurement reviewer / buyer-side governance

🌐 AEO consumer / spec implementer

🗃️ Data team

🛡️ SRE / Platform reliability stack

🤖 MCP / Claude integrator

🚦 Runtime enforcement — turning "buyer signed off" into "request denied"

The five cross-ecosystem hooks

🔌 One MCP server. 75 tools. Twelve specs + DefenseTech.

🖼️ One visualizer. Eleven specs.

⚙️ Suite governance Actions — 9-action CI ecosystem

Per-protocol fleet-summary gates (4)

Cross-protocol governance gates (3)

Cost + cluster gates (2 — pre-existing)

🛡️ Testing artifact

🌐 Live properties — 32 total

Hubs + tools (8)

Per-spec landings (11)

Earlier product surfaces (5)

Cloud Identity, Platform, FinOps & Threat Detection Governance lane (8)

🪜 Quickstart paths

📦 AEO Reference Stack (the depth example, 5 layers)

🤝 Status & contribution

📜 License

👤 Author

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors