What this is

The shortest path to a real diagnostic: authenticate, call /diagnose, and read the binding constraint + the executive surface methods.

Prerequisites

• Access to a Performix workspace. Production is auth-protected (https://performix-one.vercel.app) via Clerk; preview deploys are Vercel-auth-protected.
• For local dev: npm run dev → http://localhost:3000.

1. Run the CAMS diagnostic

# POST /diagnose — runs real psychometrics against the live toolbox `reincarnation` spoke
curl -X POST https://<base>/diagnose \
  -H "Content-Type: application/json" \
  -H "<Clerk session/authorization>" \
  -d '{ "teamId": "team_123", "responses": [...] }'
# → CAMS scores + the single binding constraint that caps performance

The diagnostic is real in prod (PFX-7 HTTP + PFX-30 MCP); a mock remains only as a local-dev safety-net fallback. UI: /diagnose.

2. Executive surface methods (per team)

POST /api/teams/{teamId}/executive-readout        # framed competency-rollup + perception-gap
POST /api/teams/{teamId}/interpretability-rules   # "why this nudge / why this flag"
POST /api/teams/{teamId}/conformal-gating         # decide vs. abstain on prediction uncertainty

MCP twins: executive_readout · interpretability_rules · conformal_gating.

Honest scope

The CAMS diagnostic is wired end-to-end and live-validated (2026-06-14); some surfaces (e.g. job-spec-authoring) are scaffold. Trust AGENTS.md's wiring table for what's real vs. scaffold, and note Performix consumes toolbox/Principia services — those must be reachable.

See also

Concepts · APIs · Architecture