6SNT-RADIO

asset-sourcing

Provision REAL external data/assets automatically (a fetch script wired to the build, idempotent, with license and declared fallback), never placeholders. Any domain. Use it when the product needs an external source.

brand-evaluation-rubric

Grade visual identity with two A-F scores (Design Score / AI-Slop) and a 10 anti-pattern detector — makes principle P-03 measurable. Use it in 03 (closing BRANDBOOK) and 08 (QA gate, on the real app).

critic

Adversarial pre-gate review. Attack a deliverable to surface contradictions, blind spots, over-claims, missing cases and spec drift before the human gate — external review, never self-congratulation. Used by 02 (RESEARCH), 08 (QA) and 09 (Security) before their gate, or by the lead at any gate.

project-cleanup

For the Technical Writer (10). Inventory and delete (only with an orchestrator-approved list) obsolete root files; no archiving, no quarantine, no 'just in case'. Use it for active root cleanup.

qa-eyes

For QA (08). Drive the REAL app (computer-use on desktop / Chrome on web), capture states and catch placeholders/mocks disguised as real (P-11 audit). Use it at the QA gate, on the real app, not the mockup.

smoke-test

'Done' includes an artifact that STARTS: an automated headless smoke with ONE command, pass/fail by exit code, and tests/smoke-result.json for the release-gate DoD. Use it to author/run the smoke (08 QA) on any stack.

technical-documentation

For the Technical Writer (10). Document reality (not the ideal) for two audiences (PO and AI); if code and docs differ, code wins; update before creating; less doc than product. Use it when closing stages.

security-audit

For Security (09). READ-ONLY audit as if the repo were public: credentials, inputs, auth, logging, dependencies; local vs public-distribution severity; justified false positives. Use it before any release.

adopt-ground

Brownfield grounding + safety pass: pack an existing repo into an AI-readable inventory (Repomix, token-budgeted), scan for secrets FIRST (Gitleaks/TruffleHog or a built-in heuristic), and audit dependencies/licenses — so reconstructed specs are grounded in real code, never in memory. Used by /adopt before any artifact is reconstructed.

agent-authoring

Write or edit lean, bloat-free agent definitions (one responsibility, one deliverable, without repeating the global principles). Use it only when creating or modifying a studio agent.

architecture-decision

For the Architect (04). Evaluate the stack from the problem (not from familiarity) with 2-3 options per layer, viability criteria, and the ARCHITECTURE.md structure. Use it at the architecture gate.

brand-evaluation

For UX/UI (03). Evaluate visual proposals against rejection/approval criteria before showing them to the PO — nothing generic or template-like, its own personality. Use it when generating and iterating visual directions.

data-integrity

For the Data Modeler (05). Design data with integrity, growth, backup and migrations with rollback; access layer, single writer, read-only readers. Use it when modeling entities and the DATAMODEL.md.

diagrams-gate

Generate diagrams (C4 / mermaid) from spec or code at the architecture/data gate (04/05) so assumptions are named, not slipped through; the gate requires diagrams present and valid.

env-reproducible

For the Architect (04). Pin the runtime PER-PROJECT (mise/Volta) + engines + engine-strict + a preinstall guard + a single build command, so a global change never breaks other apps.

orchestrator-bridge

For the orchestrator. File-based dispatch-notify loop (claude-heartbeat + AMQ) without claude -p: read the disk outbox and react; each gate's approval still belongs to the PO (P-05).

qa-visual-criteria

For QA (08). Gate criteria across three dimensions (functional / visual / performance) with 0-warnings acceptance, blocking/warning classification and the QA_REPORT.md structure. Use it when running the quality gate.

research-methodology

For the Researcher (02). Research the external context from the brief, reuse before building (plugins/MCPs/APIs/libraries), with an explicit recommendation and qualified sources. Use it when building the RESEARCH.md.

security-gate

For Security (09). Real audit with SecOpsAgentKit (Gitleaks/Semgrep/Bandit/Trivy/Grype/Checkov) + a reviewdog baseline that learns from false positives; on Electron, validate isolation (contextIsolation/sandbox/CSP).

setup

Activates the CA6SNT studio in a project. Runs the studio-init scaffolder (seeds .claude/settings.json, CLAUDE.md, evals/ and the smoke) and verifies activation. Use it when installing or starting the studio in a project folder, before /startsnt.

visualizer

Generate live, self-contained HTML previews of UI and diagrams that consume the project's TOKENS.css (zero hardcoded hex, no CDN, all states, labeled placeholders). For 03 (UX/UI), 05 (Data Modeler), 06 (Frontend) and 08 (QA).

drift-protocol

For the orchestrator. Approved-vs-built protocol (Divergence): if the implementation must deviate from what was approved, stop, describe the conflict, wait for PO approval, and update the doc before coding.

qa-verification

For QA (08). 'Done is done' verification before the gate (Jenny: impl-vs-spec; Karen: reality-check that cuts stubs; task-completion-validator: end-to-end, not just compiles). Combine with qa-eyes and the anti-placeholder audit.