HELM AI Kernel OSS Traction Plan

This page keeps public launch and community work tied to source-backed HELM AI Kernel behavior. The goal is not maximum attention; it is maximum verified Kernel adoption.

Primary objective: drive verified adoption of HELM AI Kernel as the local-first fail-closed execution boundary for AI agents.

Primary conversion: visitor -> install -> run boundary -> trigger DENY or ESCALATE -> inspect signed receipt -> verify EvidencePack offline -> star, follow, discuss, or contribute.

Adoption Path

flowchart LR
    Visitor["Visitor"] --> Install["Install or clone"]
    Install --> Boundary["Run local boundary"]
    Boundary --> Verdict["Trigger DENY or ESCALATE"]
    Verdict --> Receipt["Inspect signed receipt"]
    Receipt --> EvidencePack["Verify EvidencePack offline"]
    EvidencePack --> Community["Star, discuss, or contribute"]

flowchart LR
    Visitor["Visitor"] --> Install["Install or clone"]
    Install --> Boundary["Run local boundary"]
    Boundary --> Verdict["Trigger DENY or ESCALATE"]
    Verdict --> Receipt["Inspect signed receipt"]
    Receipt --> EvidencePack["Verify EvidencePack offline"]
    EvidencePack --> Community["Star, discuss, or contribute"]

Canonical Positioning

• Public OSS name: HELM AI Kernel.
• Repository: Mindburn-Labs/helm-ai-kernel.
• Binary: helm-ai-kernel.
• Disambiguation: Mindburn Labs' HELM execution kernel for AI agents, not the Kubernetes package manager.
• Short description: Fail-closed execution firewall for AI agents: quarantine MCP tools, proxy OpenAI-compatible requests, emit signed receipts, and verify EvidencePacks offline.
• Proof vocabulary: signed receipts, EvidencePack, ProofGraph, offline verification, ALLOW, DENY, ESCALATE.

Avoid stale product-family names, generic AI governance platform language, hosted-control-plane claims for Kernel, certification claims, or Enterprise features framed as Kernel features.

GitHub Metadata

Repository topics should spend the 20 slots on the execution-boundary wedge:

ai-agents
agent-security
mcp
model-context-protocol
tool-calling
execution-firewall
ai-security
llm-security
developer-tools
self-hosted
open-source
policy-engine
zero-trust
sandbox
signed-receipts
cryptographic-receipts
evidencepack
openai-compatible
llmops
security

Do not use saas, compliance, ai-governance, generative-ai, agentic-ai, helm, openai, or proof-receipts as OSS repo topics.

README First Viewport

The README must lead with mechanism and proof:

# HELM AI Kernel

HELM AI Kernel is the fail-closed execution firewall for AI agents.

Mindburn Labs' HELM execution kernel for AI agents, not the Kubernetes package manager.

HELM evaluates agent execution requests and records every ALLOW / DENY / ESCALATE decision as verifiable evidence.

The star CTA belongs after the local proof path:

Star HELM AI Kernel if you want to follow fail-closed AI agent execution, MCP quarantine, signed receipts, and offline-verifiable EvidencePacks.

Proof Assets

Public proof assets are executable artifacts, not just polished images. Each asset must name:

• asset name
• command to generate
• expected verdict: ALLOW, DENY, or ESCALATE
• receipt path
• EvidencePack path
• offline verification command
• expected verification output
• screenshot or GIF source
• last verified commit SHA

Canonical demo ladder:

1. Unknown MCP tool enters quarantine.
2. Sensitive action returns DENY or ESCALATE.
3. Signed receipt is produced.
4. EvidencePack verifies offline.
5. Tampered receipt fails verification.

Current assets:

• MCP quarantine proof board source: helm-mcp-quarantine-demo.svg
• Sanitized transcripts: examples/launch/assets

Launch Sequence

Gate 1: GitHub readiness. README, repo description, topics, social preview, release assets, proof demos, docs, issue templates, security policy, and Discussions are source-backed and coherent.

Gate 2: Technical launch. Use Show HN and targeted security/devtools Reddit only after the local proof path is clean.

Gate 3: Ecosystem integration. Submit upstream fixtures and listings only after the repo has concrete MCP, proxy, receipt, and EvidencePack examples.

Gate 4: Broader social. Use LinkedIn, X, Product Hunt, DEV, and Hashnode only after technical audiences have a proof artifact to reference.

Show HN title: Show HN: HELM AI Kernel, a fail-closed execution firewall for AI agents

Short description: HELM AI Kernel quarantines unknown MCP tools before dispatch, governs OpenAI-compatible requests through ALLOW, DENY, and ESCALATE decisions, emits signed receipts, and verifies EvidencePacks offline.

UTM Links

Discussions

Use Discussions for high-signal participation:

• Announcements
• Proof demo help
• MCP server quarantine proposals
• Receipt and EvidencePack schema questions
• Integration ideas
• First contribution help
• Show-and-tell

Triage flow: Discussion -> maintainer triage -> reproduction, fixture, or spec -> issue.

Contributor Funnel

OSS-safe contribution streams:

• docs: quickstart clarity, troubleshooting, examples
• mcp: quarantine fixtures, server metadata, authorization examples
• proxy: OpenAI-compatible client examples
• receipts: offline verification examples, tamper tests
• evidence: EvidencePack examples
• sdk: small SDK polish and samples
• security: negative tests and threat-model examples
• ecosystem: factual integrations and listings

Every contributor issue should include context, exact file paths, expected output, validation command, acceptance criteria, and out-of-scope boundaries. Do not expose Enterprise implementation details in OSS issues.

Measurement

Awareness:

• GitHub visitors, referrers, and popular content
• social, Hacker News, Reddit, and docs traffic

Evaluation:

• README and quickstart visits
• Homebrew install interest or available formula analytics
• clone count
• release asset downloads

Proof:

• boundary started
• proof demo run
• DENY or ESCALATE receipt generated
• EvidencePack exported
• EvidencePack verified offline

Engagement:

• stars
• watchers
• discussions
• issues
• PRs
• ecosystem mentions

Commercial bridge:

• HELM AI Enterprise Individual interest CTA clicks
• docs path from Kernel to Individual
• inbound team-use requests

GitHub traffic windows are short, so export private analytics daily, publish a weekly channel cohort report, keep a launch-post UTM map, and log README hook changes.

Truth Gate

Run this checklist before publishing README edits, launch posts, diagrams, videos, social previews, and website copy:

• Uses HELM AI Kernel as the current public name.
• Mentions not Kubernetes Helm when context is introductory.
• Uses signed receipts, EvidencePack, and ProofGraph exactly.
• Uses ALLOW, DENY, and ESCALATE.
• Does not claim seccomp or eBPF unless current code proves it.
• Does not claim a hosted control plane for OSS Kernel.
• Does not claim certification unless certification exists.
• Does not describe Enterprise features as Kernel features.
• Does not imply Individual or Enterprise weakens or forks Kernel semantics.
• Does not mention robot fleets, AGI OS, OrgDNA compiler, Titan, or physical-world control in OSS copy.

Diagram Doctrine

OSS hero diagrams should use this shape:

Agent / Model / Orchestrator
  -> Action Proposal
  -> HELM AI Kernel Boundary
  -> ALLOW / DENY / ESCALATE
  -> Signed Receipt
  -> ProofGraph
  -> EvidencePack
  -> Offline Verification

Keep CompanyArtifactGraph, GeneratedSpec, and Enterprise loop diagrams on Enterprise bridge pages, not the OSS hero.

Ecosystem Rule

No listing PR without a working integration, fixture, or reproducible example. Prioritize:

• MCP quarantine fixture
• OpenAI-compatible proxy example
• receipt verification fixture
• EvidencePack tamper-negative test
• policy DENY or ESCALATE example

Submit listings after the proof exists.

Commercial Bridge

Use this bridge after local Kernel proof:

Try HELM AI Kernel locally.
Use HELM AI Enterprise Individual for shared approvals, receipts, policies, and short-retention evidence.
Talk to Mindburn Labs about production execution authority.

HELM AI Enterprise Individual gives operators a shared control plane for governed AI actions: workspaces, approvals, receipts, API access, custom policies, and short-retention evidence, all built on HELM AI Kernel.

Source Truth

• Quickstart
• Developer journey
• Execution security model
• Publishing runbook
• Version surfaces contract