Pick a framework. Eight engines reshape every artifact downstream.
Every framework you tick on Tuttify Carbon's Smart Intake doesn't just decorate the cover page. It propagates through eight downstream engines — biasing intensity dials, swapping in framework-specific agents, spawning new artifacts, and forcing override locks when the cost of getting it wrong includes criminal liability.
Every framework that puts you in front of an auditor — pre-wired.
The catalog grew from 12 to 15 with the addition of iso_42001, nist_ai_rmf, and the split of fda_samd from HIPAA. Every cue carries documented downstream effects — including its strictness rank and intensity bias.
Export controls. Forces safety_intensity high and thorough gate visibility — cannot be silently weakened.
Criminal Justice Information Services. Forced override engages automatically.
Control families AC · AU · CM · CP · IA · IR · RA · SC · SI · CA. Default template targets FedRAMP Moderate; High and IL5 are configurable. Artifacts: SSP, POA&M, CRM, ConMon plan, SAR-ready evidence stubs by family.
PHI handling sections, BAA scope, audit-log retention horizons (164.312(b)), breach-response artifacts (164.408 / 164.410). Control citations to 164.308–164.316 throughout.
IEC 62304 software lifecycle (Class A/B/C). ISO 14971 risk management — hazards, harms, probability, severity. 21 CFR Part 820 design controls. 2023 FDA Cybersecurity Premarket Guidance + SBOM. Risk register switches to 14971 structure; artifacts shape for a 510(k) DHF.
CSF v11 control specifications mapped across the 19 categories. Inheritance-tier reasoning (System / Customer / Vendor). MyCSF-shaped evidence stubs and r2-assessment scope statement.
Cardholder Data Environment diagram, segregation-of-duties tables, change-control records.
Safeguards rule mapping, NPI inventory, vendor-risk artifacts.
Spawns a dedicated ai-governance.md artifact. AIMS scope, AI-impact assessment, lifecycle responsibilities.
Govern · Map · Measure · Manage tagging on every AI-touching requirement.
TSC mapping for Security, Availability, Confidentiality and Privacy. Per-control narrative stubs (CC6.1 access control, CC7.2 monitoring, etc.) keyed to your system description.
Article 30 records, DPIA triggers, data-subject rights flow.
Consumer rights flows, opt-out plumbing, sensitive PI tagging.
Student-record protection, parental consent flow.
A11y acceptance criteria injected into every UI requirement.
The synthetic baseline. Carbon still ships PRD, SRD, threat model — without framework-specific overlays.
Every checkbox propagates. Nothing is hand-waved.
Every flag you select is wired through a documented effect map. Nothing implicit. Nothing hand-waved. You can audit the bias map in source as derived_defaults.py::_REGULATORY_INTENSITY_BIAS.
Validates the cue against the project shape. Surfaces missing intake answers required by the framework.
Computes the project profile: gate visibility, tone register, RQS weighting, audit cadence.
Picks the agent roster. A HITRUST run brings 8+ agents. A landing page brings 4.
Sets compliance_intensity and security_intensity independently. Each acts as a floor — never weaker than baseline.
Modifies compliance-risks.md and risk-register.csv. May spawn entirely new artifacts (e.g. ai-governance.md).
Adds attacker payload sections and framework-specific abuse cases to threat-model.md.
Pulls the framework's required section structure into the output template. No ad-hoc layouts.
For Rank-5+ frameworks: locks safety_intensity=high and thorough gate visibility. Cannot be silently weakened.
Compliance and Security don't share a dial.
The legacy single safety dial was split deliberately. A FedRAMP build needs maximal security intensity — but a Section 508 + GDPR run needs compliance intensity dialed up without the threat model getting heavier than the scope warrants.
Each framework you tick carries a documented bias on both dials. The biases stack. The result is the higher floor — never weaker than any single framework would have demanded.
Floors only · stacked frameworks raise the floor, never lower it.
For criminal-penalty frameworks, there is no soft path.
If your build touches HIPAA, FDA SaMD, FedRAMP/NIST, HITRUST, ITAR/EAR or CJIS, Carbon forces safety_intensity=high and the Intervention Budget to thorough — every gate visible. This override cannot be silently weakened by a less-strict base rule elsewhere in the cue mix.
- ✗A lighter framework in the mix downgrading the FedRAMP floor.
- ✗A "fast" run hiding G3 or G4 review for a HIPAA build.
- ✗A silent demotion of intensity by another engine downstream.
The reviewer doesn't get a doc. They get receipts.
Provenance pills are first-class UI. Every requirement, every framework citation, every diagram traces to the intake answer, supporting document, or generated artifact that produced it.
Every state-changing event is appended to a hash-chained log. carbon audit verify greens the chain. The tamper-demo reds it. Reviewers see the seal.
Deterministic. Not LLM-graded. EARS pattern detection + INCOSE rule checking. The reference demo lifts from 0.82 to 0.93 after one G4 refinement — visible, audited, repeatable.
Bring us your hardest regulatory cue mix.
FedRAMP + HIPAA + ISO 42001 in the same build? That's a working session. We'll run it through Carbon and walk you through the bundle that comes out.