Basaltborne

Mission

Build a public publishing platform where drafts cannot become public because a UI flag flipped, and where AI-authored changes cannot reach main without tests and review.

Problem

AI coding agents can produce a lot of surface area quickly. Without phase gates, RLS, and an independent CI suite, that speed becomes a way to ship unauthorized data and unreviewed behavior.

Constraints

• AI output is treated as untrusted until validated.
• Advancing a phase requires acceptance criteria, passing tests, human review, source authorization, and explicit security boundaries.
• Privileged service-role credentials are never shipped to the browser.
• The browser is untrusted by default.

My role

I direct AI coding agents through a phase-gated lifecycle: they can research, propose architecture, implement, write tests, document, and troubleshoot — they cannot advance a phase or publish on their own.

Architecture

Next.js / React / TypeScript public and admin surfaces; PostgreSQL / Supabase with Row-Level Security; versioned migrations; API/RPC boundaries; Zod at trust boundaries; GitHub Actions running Vitest, pgTAP, Playwright, lint/type gates, and migration replay.

Build process

1. Phase 1: repository, database, authz, approval/audit concepts, admin scaffold, migrations, testing baseline.
2. Phase 2: public site, navigation, responsive design, SEO, accessibility, methodology / affiliate / corrections pages as first-class content.
3. Phase 3: draft versus approved/published content; immutable versioned publication; public-read boundaries; provenance; auditability.

Validation

Lessons

• Publication is privileged

  Publication is treated as a privileged state transition enforced server-side — not a boolean flag flipped in the UI.

• Branch protection is an approval control

  The pull request is where automated validation and human judgment both have to sign off before anything reaches main.

Cybersecurity equivalent

• Row-Level Securitymaps toAuthorization enforced at the data layer
• Server-only service rolemaps toCredential protection / no secret in the client
• CI as independent checksmaps toPreventive and detective controls on change
• Human-approved deploymaps toPrivileged change authorization

AI principles applied

• Human-in-the-loop

  Consequential actions require a person. The loop is a state machine, not a polite request in a prompt.

  Where learned: Publishing, packaging, containment, and deploy all use explicit gates.

  See in the AI map →

• Trusted retrieval / source timing

  Normalize and timestamp sources before the model sees them. Forum consensus is not a specification.

  Where learned: TinyClaw freshness path; Jeep source-verification instruction.

  See in the AI map →

• Validation and evaluation

  Happy path is not enough. Missing, malformed, stale, contradictory, injected, and timed-out inputs are the real suite.

  Where learned: Triage functional plus adversarial layers; Clearward 49 automated checks; Basaltborne CI layers.

  See in the AI map →

• Authority boundaries

  Collection, analysis, and remediation are different privileges. Drafting and publishing are different privileges.

  Where learned: Triage and Basaltborne both make that separation structural.

  See in the AI map →

• AI-assisted development

  Agents draft under persistent instructions and phase gates. They do not merge, package, or deploy themselves.

  Where learned: Clearward AGENTS.md; Basaltborne phase-gated lifecycle with PR review of AI-authored changes.

  See in the AI map →

Security principles applied

• IAM / authorization

  Who can do what is assigned outside the model and enforced at the boundary that matters.

  Where learned: Basaltborne RLS and RPCs; Clearward human-only credential boundary; triage least-privilege account.

  See in the security map →

• Secure SDLC

  Phase gates, PR review, protected main, and tests that run somewhere other than a laptop.

  Where learned: Basaltborne feature branches, acceptance-criteria commits, CI-protected main.

  See in the security map →

• Logging and accountability

  Decisions, tool calls, retries, and approvals have to be reconstructable.

  Where learned: Triage adversarial study scored whether attempts appeared in the audit log.

  See in the security map →

• Validation

  Schema, hashes, required fields, RLS tests, and visual review — not the generator marking its own homework.

  Where learned: Clearward formula/package QA; Basaltborne Zod + pgTAP; TinyClaw source checks.

  See in the security map →

• Change control

  Publication, packaging, and deploy are privileged transitions with a human on the gate.

  Where learned: Clearward release architecture; Basaltborne draft vs published; Self Help human review.

  See in the security map →

• Secrets management

  Credentials are a human-only boundary. Service roles stay server-side. Prompts are not a vault.

  Where learned: Clearward AGENTS.md; Basaltborne server-only privilege.

  See in the security map →

• Incident response / recovery

  Recommended actions include expected results, validation, operational risk, and rollback. Checkpoints revert test cycles.

  Where learned: Triage recommendations and Hyper-V checkpoints; Basaltborne planned undo for migrations.

  See in the security map →

• Application security

  Untrusted browser, validated RPCs, private tables, fail closed, information minimization.

  Where learned: Basaltborne database-security and public-API design.

  See in the security map →

Technologies

• Next.js
• React
• TypeScript
• PostgreSQL
• Supabase
• Zod
• GitHub Actions
• Playwright
• Vitest
• pgTAP
• ESLint
• Docker
• PowerShell

Skills demonstrated

• Systems Architecture
• Next.js
• TypeScript
• React
• PostgreSQL
• Supabase
• Row-Level Security
• Database Migrations
• Schema Validation (Zod)
• CI/CD
• Unit / Integration / E2E Testing
• Database-Security Testing (pgTAP)