Generated reference: edit sources under docs/helix/, then npm run site:build.

Generated from HELIX docs. Source: docs/helix/01-frame/prd.md. Edit the source, not this file or site/ HTML.

ddx:
  id: aibadge.prd
  type: prd
  kind: product
  links:
    - target: aibadge.product-vision
      kind: informed_by
    - target: aibadge.competitive-analysis
      kind: informed_by
    - target: aibadge.research.prior-art
      kind: informed_by
    - target: aibadge.research.naming
      kind: informed_by
    - target: aibadge.principles
      kind: informs
    - target: aibadge.concerns
      kind: informs
status: draft
product_name: Innsigle
activity: 01-frame
created: 2026-07-22
updated: 2026-07-25

Product Requirements Document

Summary

Innsigle is the maker's seal for published work—a content-origin seal for makers (publishers/operators) who publish in two channels. Category: content provenance for the AI era.

1. UC-AI-docs: Owned documentation (e.g. HELIX microsite): model-primary

content with a signed colophon that can proudly name tools (e.g. Claude) and a house/person signer.

1. UC-human-social: Short posts on X/Twitter and similar: a human-authored

mark that stays visible when platforms strip file metadata.

Same seal family; different colophons. Optional DKIM-style signing for durable content. Not an AI detector. Not a C2PA replacement (C2PA/SynthID serve media pipelines and vendor watermarks; social text and proud model BOM are the gap).

v1: Mark system + colophon model + docs sign/verify CLI + social placement pattern (image/link). Sample on operator docs and social first. Editorial tooling such as sloptimizer (easel-skills) is a complementary pipeline step: list it on the bill when used; it does not change composition to human-authored.

Name: Innsigle (INN-siggle, rhymes with single).

Problem and Goals

Problem

Two concrete failures for the same operator:

Competitive layer map (see competitive-analysis.md): badges, C2PA, watermarks, IPTC codes, platform labels, and detectors each cover a slice; none cover both UCs as one brand + claim system.

Goals

1. UC-AI-docs: Operator can publish docs with an Innsigle whose colophon

names generative tools and is optionally signed so readers verify house/person sealed that bill for those content bytes.

1. UC-human-social: Operator can mark a short social post as human-authored

with the same seal family, without depending on embedded file manifests.

1. Product copy never markets detection or treats AI composition as a lesser

class of seal.

Success Metrics

Non-Goals

• AI content detection or decoding SynthID/Video Seal.
• Replacing C2PA/CAI as the media industry standard or camera capture chain.
• Guaranteeing colophon factual truth (only declaration + optional signature).
• Multi-tenant accounts, marketplace, or social network in v1.
• Native X API integration or platform partnership as P0.
• EU AI Act certification product.
• Requiring signed claims on ephemeral social posts.
• Treating prose de-AI-ifying (e.g. sloptimizer rewrite) as proof of

human-authored composition or as an authorship detector.

• Absorbing sloptimizer into Innsigle; they remain separate tools that can run

in sequence.

Users and Scope

Primary Persona: Dual-channel operator (Erik / Azgaard)

Role: Builds methodology docs with models; posts human writing on social Goals: Proud model BOM on docs; clear human mark on social; one system Pain: Tools force either human-only badge culture or industrial media provenance

Secondary Persona: Docs reader

Role: Developer or practitioner reading sealed documentation Goals: See which models/tools produced the page; optional verify signer Pain: Unclear whether docs are human, model, or mixed; no trust anchor

Tertiary Persona: Social viewer

Role: Follower on X/similar Goals: Tell human-authored posts from unlabeled AI sludge at a glance Pain: Platform labels incomplete; metadata invisible

Requirements

Must Have (P0)

1. Colophon model: at least human-authored, model-primary (AI-generated), and

mixed; tools/models nameable (e.g. Claude).

1. Visual seal family usable for both composition poles (not separate good/bad logos).
2. UC-AI-docs: Embed mark + colophon on a static/HTML docs page; optional

sign and verify for canonical content.

1. UC-human-social: Placement pattern for human-authored posts when file

metadata cannot be trusted (mark image and/or short verify/profile link).

1. Public "what this is not" (not detector; not C2PA replacement).
2. Runnable sign/verify tooling for the docs path (CLI or equivalent).

Should Have (P1)

1. Brand/docs site teaching both use cases in ≤5 minutes.
2. House signer profile (Azgaard) as default sample issuer.
3. Site-generator notes (e.g. Hugo/static footer partial).
4. Social asset pack (avatar-adjacent mark, post image template, caption templates).
5. Documented sample recipe: model draft → sloptimizer → Innsigle bill/sign for

methodology docs (links to easel-skills sloptimizer; no merge of codebases).

Nice to Have (P2)

1. C2PA export/bridge for image assets.
2. IPTC digitalSourceType mapping from colophon fields.
3. Hosted key directory.
4. Native platform integrations.
5. Quantitative percentages in colophon.
6. Optional agent/skill hook that proposes a colophon after a sloptimizer pass

(still human-confirmed composition state).

Functional Requirements

Subsystem: Colophon Model

• FR-1 Colophon includes composition states at least: human-authored,

model-primary (AI-generated), mixed.

• FR-2 Colophon can list named tools/models (e.g. Claude, other LLMs,

image models), editorial tools (e.g. sloptimizer), and human roles (edit, structure, review).

• FR-3 No required ranking of states as better/worse; AI-named bills are

valid first-class seals (model credit is an allowed primary use).

• FR-4 Product defaults and mark variants do not encode human = good /

AI = bad.

• FR-4a Composition state is independent of prose "AI-tell" cleanliness:

an editorial rewrite (including sloptimizer) must not, by itself, flip model-primary or mixed to human-authored. Guidelines state this boundary.

Subsystem: Visual Seal System

• FR-5 Seal assets form one visual family across composition states; variants

may signal state without purity/shame coding.

• FR-6 Placement guidelines cover (a) docs site footer / about, (b) social

mark image or link pattern.

• FR-7 Guidelines forbid "verified real content" / oracle-truth motifs.

Subsystem: UC-AI-docs (documentation)

• FR-8 Maker can publish an Innsigle on an HTML/static docs page with colo

inline or one link away, including named models.

• FR-9 Maker can sign a claim over a defined canonical content unit (page or

artifact) binding: content fingerprint, colophon, issuer, time.

• FR-10 Third party verifies claim + content + issuer public key; fails

closed on mutation or key mismatch.

• FR-11 v1 signing uses maker-controlled local keys; no hosted account required.
• FR-12 Verify output reports validity and signer, not a detection verdict.

Subsystem: UC-human-social

• FR-13 Maker can mark a short social post as human-authored using assets

or patterns that do not require surviving C2PA/EXIF on the platform.

• FR-14 Social pattern includes a path for viewers to learn mark meaning

(profile link, site, or short URL); unsigned is allowed.

• FR-15 Social human mark uses the same seal family as docs (not a separate

human-only brand product).

Subsystem: Brand Site and Education

• FR-16 Public explainer documents both UCs, colophon examples (Claude-named

docs, optionally with sloptimizer on the bill; human social), verify when signed, non-goals, and the sloptimizer complement (edit vs seal; no authorship laundering). Brand lines (byte-exact, straight '): home H1 and footer chrome use The maker's seal for published work; category line Content provenance for the AI era on Non-goals/meta—not as sole footer pitch. Seal/mark ≠ signet (unsigned seals valid). Maker = publisher/operator. Sigil at most once in home or Marks body plus glossary; no occult register.

• FR-17 Pronunciation documented once: INN-siggle (rhymes with single).

Subsystem: Tooling Distribution

• FR-18 Assert/sign/verify tooling ships documented and runnable for UC-AI-docs

P0 without a GUI.

• FR-19 Tooling accepts a colophon that includes named generative tools.

Acceptance Test Sketches

Technical Context

Assumptions until ADRs:

Constraints, Assumptions, Dependencies

Constraints

Assumptions

• Naming models (Claude, etc.) in a public bill is desirable for UC-AI-docs.
• Naming editorial tools (sloptimizer) on the bill is desirable when used.
• Human social often stays unsigned; in v1, visible mark placement matters more than per-post signatures.
• Local keys acceptable for house signing of docs.
• Sloptimizer remains in easel-skills; Innsigle does not reimplement Vale prose rules.
• Competitive analysis in docs/helix/00-discover/competitive-analysis.md remains the rivalry baseline until next research pass.

Dependencies

• Competitive analysis and vision (this pass)
• Seal mark design (open)
• Claim format ADR/contract (open)
• Domain for verify URLs (open)
• Optional: easel-skills sloptimizer available in operator agent environment for sample recipe

Risks

Open Questions

Design drafts: docs/helix/02-design/ (DESIGN.md, claim-system.md, attestation-prior-art.md). Remaining freezes:

• [ ] Canonical content unit for multi-page docs (per-page vs release set)
• [ ] HTML canonicalization for hashing (Contract)
• [ ] Signature algorithm + envelope freeze (ADR; draft proposes Ed25519)
• [ ] Seal glyph geometry and state cues (DESIGN.md open list)
• [ ] Product domain vs *.azgaard.net for type URIs and key_url
• [ ] Model version pins in colophon v1
• [ ] Claim URL shape for social (profile well-known vs per-post)

Success Criteria

• UC-AI-docs live on at least one operator docs property with named-model colophon;

signed verify path green on one canonical artifact.

• UC-human-social live as at least one real human-authored social placement.
• Automated sign/verify tests green; no detection claims in public materials.
• Competitive non-goals held: not C2PA replacement, not detector.

Review Checklist

• [x] Summary is a 1-pager including both UCs
• [x] Problem tied to competitive gaps
• [x] Goals map to UC-AI-docs and UC-human-social
• [x] Metrics cover live samples on both channels
• [x] FR-n cover colophon, docs sign, social pattern
• [x] Non-goals exclude C2PA war and detectors
• [x] Acceptance sketches cover both UCs