Claude Code
Plugin commands plus six eval cases for natural-language use and safety behavior.
Deterministic process for capable agents.
Give capable agents a process they cannot improvise.
Tests passing831
Subtests passing2,849
Runtime dependencies0
Signed release packs4
The refusal is the product.
Most agent tooling asks for trust and returns prose. Exemplar returns bounded work: one route, declared inputs, approval gates, deterministic plans, cited evidence, and signed bytes. Missing evidence never becomes invented output.
The collection
Four packs. One standard of proof.
Each download includes the signed pack, the Exemplar CLI wheel, a plain-language operator guide, real samples, license terms, and the public key needed to verify what you received.
01 · Flagship
Marketing that knows which workflow comes next.
01Research to strategypositioning
02SEO performance planevidence gated
03Content enginedistribution
04Email lifecycleretention
05Paid amplificationconversion first
18 workflows · 18 capabilities585 indexed prompts
Exemplar Marketing
One marketing goal in. One exact workflow out.
Routes plain-language goals across strategy, SEO, content, email, paid distribution, local growth, and delivery. The pack never invents rankings, metrics, sources, or schedules.
- 65 authored routes
- 18 workflows and capabilities
- 12 playbooks, six audience profiles
- 585-entry prompt index
$250one time · solo
Buy Marketing →02 · Legal operations
NDA review with a hard boundary.
Definition of confidential informationexcerpt cited
Permitted use and disclosureexcerpt cited
Exclusions from confidentialityexcerpt cited
Term and survivalexcerpt cited
Return or destructionexcerpt cited
Out-of-scope contracts stop before review. Attorney approval stays required.
Exemplar Legal Ops
Useful because it knows what it cannot do.
Triage the request, assess supplied NDA excerpts against an authored 11-point checklist, and produce an advisory redline checklist for attorney review.
- NDA-only scope guard
- Every finding cites its excerpt
- Attorney-review disclaimer by construction
- Word-ready memo export
$99one time · solo
Buy Legal Ops →03 · AI quality
One readiness record across tests and evals.
conventional testsnormalized
JUnit evidence record
Source-linked pass, failure, and skip evidence.
agent evaluationsnormalized
Evaluation evidence record
Authored criteria, observed failures, and safety findings.
release readinesshuman decision
Advisory synthesis
Evidence gathered. Release approval remains outside the pack.
Exemplar AI QA
Evidence before the go or no-go meeting.
Normalize conventional test results and AI agent evals into one advisory readiness record. Safety findings are retained and cannot be waived by the pack.
- Conventional and agent assessments
- Evidence-cited synthesis
- Three readiness profiles
- Never approves a release
$99one time · solo
Buy AI QA →04 · Recruiting
Candidate screening that cites the record.
criterionstatussource
Relevant domain workmetprofile §2
Required operating scopemetprofile §4
Team leadershipunverifiedno source
Location requirementmetprofile §1
Final decisionhumanoutside pack
Exemplar Recruiting
A consistent process, not a claim about better hires.
Turn a role brief into a schema-valid rubric, screen profiles against it, and produce an evidence-linked shortlist. Unverifiable claims stay marked as unverifiable.
- Rubric-based scorecards
- Every claim linked to source evidence
- CSV shortlist export
- Hiring decision stays human
$99one time · solo
Buy Recruiting →The full collection
Four packs, one verified operating layer.
Marketing, Legal Ops, AI QA, and Recruiting. Buy once and keep every v1 update. Save $98 against individual pack pricing.
$449one time · solo license · all sales final
Get all four →
Proof, not posture
The trust claim is machine-checkable.
The release suite ran against the exact sources used to build v0.7.0 on September 16, 2026. The same bundle can be verified offline on your machine.
ValidatePack contracts and fixtures
RebuildByte-identical bundle
VerifyEd25519 signature and hash
OperateApproval-gated execution
verification.log
$ exemplar route packs/recruiting \ "screen this batch" --compact { "status": "needs_input", "missing_inputs": [ {"name": "screening-rubric"}, {"name": "candidate-profiles"} ], "reason": "missing_required_inputs" } $ exemplar verify-bundle \ --expect-sha256 36af8c51…09381 \ pack-ai-qa-1.1.0.exemplarpkg verified: signature OK verified: sha256 match $ python3 -m pytest -q 831 passed, 10 skipped, 2849 subtests passed
Test run: September 16, 2026
Agent surfaces
Use the model you already trust.
The operating contract is portable. Exemplar ships the same verified commands and safety rules across the agent surfaces teams already use.
Codex
A focused skill wrapper exposes the same verified command surface and operating contract.
MCP
Eight JSON-RPC tools over stdio. Remote run requests return an approval-required plan.
OpenAI Agents API
Documented patterns for durable sessions, subagents, and controlled pack operations.
Licenses and service
Start self-serve. Bring the process to your team later.
Solo
Specialist pack
Legal Ops, AI QA, or Recruiting, with all v1 updates.
$99Solo
Exemplar Marketing
The flagship 65-route marketing operating system.
$250Solo
All four packs
Complete collection, one operator, save $98.
$449Team
Organization license
Up to 10 operators, all packs, one year of updates and email support.
$499/yrCustom
7-day pack build
Your workflow encoded, tested, signed, documented, and handed over.
$1,000FAQ
The practical questions.
What exactly arrives after purchase?+
A signed .exemplarpkg bundle, the Exemplar 0.7.0 CLI wheel, a registry entry with SHA-256 pins, the public verification key, a plain-language operator guide, real samples, and the license terms.
What do I need to run it?+
Python 3.11 or newer. Routing, validation, compilation, test fixtures, menus, and offline verification work immediately. Executing external work requires adapters you explicitly register and state storage you provide.
Does the pack execute work on its own?+
No. Route and compile are non-executing. The run command needs registered adapters, state storage, and explicit approval. MCP run requests return an approval-required plan rather than executing remotely.
Can I inspect the product before trusting it?+
Yes. The verification report, CLI output, samples, operator guides, signature format, and refusal cases are part of the buyer package. Every commercial bundle verifies offline.
Is this legal advice, a hiring decision, or a release approval?+
No. Legal Ops findings require attorney review. Recruiting reports rubric evidence but never claims someone is qualified. AI QA produces advisory readiness evidence and never approves a release.
What is the refund policy?+
All sales are final. These are instant digital downloads. The product scope, technical requirements, and verification process are stated before purchase.
Stop wondering whether the agent did the work.
Install the process. Verify the bytes. Keep the human decision where it belongs.