docs(portfolio): add evidence-backed conversion assets

README.md

@@ -220,6 +220,15 @@ and unsupported claims. Catalog counts are historical observations—not proof
 that every listed route is live—and quantified portfolio claims require a
 reproducible artifact before they can be presented as verified results.
 
+## Portfolio and consulting
+
+The [portfolio proof matrix](docs/portfolio/PORTFOLIO_PROOF_MATRIX.md) maps the
+project’s audience-specific stories to reproducible evidence and states the
+limits of each claim. For teams operating multiple AI gateways, the [AI Gateway
+Assurance Audit](docs/portfolio/AI_GATEWAY_ASSURANCE_AUDIT.md) describes a
+scoped review of route identity, capability qualification, policy/failover,
+and privacy-safe evidence.
+
 ---
 
 ## Project Structure

docs/portfolio/AI_GATEWAY_ASSURANCE_AUDIT.md

@@ -0,0 +1,95 @@
+# AI Gateway Assurance Audit
+
+## A practical review of what an AI route can do, may do, and can prove
+
+Verdict is built for teams that have multiple gateways, providers, model
+aliases, tools, and fallback paths but cannot confidently answer whether a
+route is actually eligible for a task. The AI Gateway Assurance Audit turns
+that uncertainty into a scoped engineering report.
+
+This is an engineering review of a configured route set. It is not a promise
+that a provider is available, that a model is intelligent, or that a system is
+production-ready.
+
+## What the audit covers
+
+1. **Inventory and identity** — enumerate configured gateways, providers,
+   connections, endpoints, protocols, model revisions, aliases, and fallback
+   edges without collapsing distinct executable routes.
+2. **Capability qualification** — review claimed metadata separately from
+   direct observations; run only consented, bounded protocol, schema, and tool
+   checks against approved test routes.
+3. **Policy and failover behavior** — inspect hard capability predicates,
+   freshness, quota/cooldown handling, fallback legality, and whether advisory
+   ranking can reintroduce an excluded candidate.
+4. **Evidence and privacy** — verify that decisions, execution outcomes, and
+   qualification reports are scoped, redacted, integrity-linked, and useful to
+   an operator without retaining prompts or response bodies.
+5. **Remediation plan** — return prioritized gaps, an owner-ready backlog,
+   requalification triggers, and a release-evidence checklist.
+
+## Inputs
+
+- A route and provider inventory, or a read-only export of the configured
+  gateway catalog.
+- The task capabilities and risk classes that must be hard requirements.
+- Optional opt-in access to a test environment for bounded probes. Live calls
+  are never implied by catalog membership or HTTP success.
+- Existing policy, CI, incident, and evidence artifacts that the customer
+  wants reviewed.
+
+Credentials and raw prompts are not required for the offline contract review.
+If live probes are authorized, credentials remain in the customer-controlled
+environment and the public report contains only sanitized diagnostics and
+digests.
+
+## Deliverables
+
+- A route identity and capability matrix showing `supported`, `unsupported`,
+  and `unknown` states with provenance and expiry.
+- A redacted qualification report explaining why each requested capability is
+  admitted or rejected.
+- A policy/failover findings register with severity, evidence reference,
+  reproduction step, remediation, and residual limitation.
+- A release-readiness checklist that distinguishes local contract evidence,
+  observed provider evidence, and evidence that is still missing.
+- A 30-minute handoff walkthrough for the engineering owner.
+
+## Typical timeline
+
+The offline review is designed as a focused one-week engineering engagement:
+
+- **Day 1:** scope, threat model, route inventory, and success criteria.
+- **Days 2–3:** contract, identity, policy, and evidence review.
+- **Days 3–4:** optional consented qualification fixtures or bounded test-route
+  probes.
+- **Day 5:** report, remediation backlog, and handoff.
+
+The timeline is a planning target, not a service-level guarantee; provider
+access, route count, and review scope determine the actual effort.
+
+## Explicit exclusions
+
+This audit does not certify provider uptime, model quality, security of a
+provider’s infrastructure, regulatory compliance, production readiness,
+adoption, performance leadership, or a specific business outcome. It does not
+publish credentials, private endpoint details, raw prompts, response bodies,
+or customer data. Any quantitative claim must have a dated, reproducible
+artifact with a defined metric, baseline, environment, and raw-result boundary.
+
+## Evidence boundary
+
+The public Verdict repository demonstrates deterministic local contracts for
+eligibility, passports, protocol qualification, structured output, tools,
+receipts, and credential-free quickstart behavior. Those contracts are not a
+substitute for customer-specific live evidence. The [public proof
+matrix](../proof/EVIDENCE_INDEX.md) and [claims ledger](../proof/claims_ledger.v1.json)
+show what is verified, observed, partial, or explicitly not approved.
+
+## Contact and next step
+
+Open a scoped request in the repository’s [GitHub issue
+tracker](https://github.com/mrnicholasbcarter-code/verdict-core/issues) with
+the route count, required capabilities, environment constraints, and desired
+deadline. Do not include credentials, raw prompts, customer data, or private
+endpoint URLs in the issue.

docs/portfolio/PORTFOLIO_PROOF_MATRIX.md

@@ -0,0 +1,37 @@
+# Portfolio proof matrix
+
+This is the recruiter, hiring-manager, and client-facing index for the
+Verdict portfolio. It keeps the conversion story tied to evidence that can be
+opened and reproduced.
+
+| Audience | Proof asset | What it demonstrates | Boundary |
+| --- | --- | --- | --- |
+| Hiring manager | [Credential-free quickstart](../../README.md#run-the-credential-free-flagship-quickstart) | A deterministic local decision flow and explicit hard-gate exclusions | It is a fixture, not live provider execution |
+| Platform engineer | [Capability passports](../CAPABILITY_PASSPORTS.md) and [qualification reports](../QUALIFICATION_REPORTS.md) | Exact route identity, provenance, expiry, redaction, and fail-closed admission | A report projects existing evidence; it does not create live evidence |
+| Reliability/security reviewer | [Public evidence index](../proof/EVIDENCE_INDEX.md) | Traceability from claims to source, tests, schemas, and limitations | Release readiness remains partial until an exact tagged bundle exists |
+| AI platform client | [AI Gateway Assurance Audit](AI_GATEWAY_ASSURANCE_AUDIT.md) | A concrete review scope, deliverables, inputs, exclusions, and handoff | Customer-specific provider findings require customer-authorized evidence |
+| Systems/quant reviewer | [Kalshi case study](KALSHI_TRADING_BOTS_CASE_STUDY.md) | Architecture and risk-control narrative | Its quantitative figures are self-reported and not independently verified here |
+
+## Approved positioning
+
+> I build trustworthy decision systems under uncertainty—from AI execution
+> assurance to capital and risk systems.
+
+> Verdict tells you what an AI stack can actually do, what it is allowed to
+> do, and what evidence proves what happened.
+
+These lines describe the engineering thesis and the repository’s evidence
+boundary. They do not claim adoption, production readiness, model quality, or
+performance leadership.
+
+## Before publishing a stronger claim
+
+Run the proof validator and inspect the linked row in the claims ledger:
+
+```bash
+python scripts/verify_proof_matrix.py
+```
+
+Do not promote a claim from `unsupported`, `self_reported`, `aspiration`, or
+`partial` until the ledger’s missing evidence is present, reproducible, dated,
+and linked to the exact released artifact under discussion.

docs/proof/EVIDENCE_INDEX.md

@@ -17,6 +17,14 @@ what it deliberately refuses to claim.
 | Adversarial checklist | [`ADVERSARIAL_REVIEW_CHECKLIST.md`](ADVERSARIAL_REVIEW_CHECKLIST.md) |
 | Validator | `python scripts/verify_proof_matrix.py` |
 
+## Conversion assets
+
+- [Portfolio proof matrix](../portfolio/PORTFOLIO_PROOF_MATRIX.md) maps each
+  audience to an evidence-backed project story and preserves limitations.
+- [AI Gateway Assurance Audit](../portfolio/AI_GATEWAY_ASSURANCE_AUDIT.md)
+  defines the scoped consulting offer, deliverables, exclusions, and safe
+  contact path.
+
 ## Verified local contracts
 
 - Hard eligibility is applied before advisory ranking; excluded candidates