Canonical report channel as L3 source of truth; blog channel MIF-exempt
ADR-0007: Canonical report channel as L3 source of truth; blog channel MIF-exempt
Section titled “ADR-0007: Canonical report channel as L3 source of truth; blog channel MIF-exempt”Status
Section titled “Status”Accepted
Context
Section titled “Context”Background and Problem Statement
Section titled “Background and Problem Statement”The harness emits research into more than one channel: a canonical markdown
report and published prose (blog, and channel packs such as book). The question
is which channel is the authoritative L3 artifact, and how a published channel is
exempted from the L3 gate without exempting it silently
(docs/explanation/mif-io-conformance.md, harness.config.json outputs[]).
Current Limitations
Section titled “Current Limitations”All channels are MIF Level-1 outputs, but holding a published format to the full L3 conformance gate is the wrong bar — a blog post’s published format is orthogonal to the structured finding graph, and embedding internal finding identity in published prose would leak it.
Decision Drivers
Section titled “Decision Drivers”Primary Decision Drivers
Section titled “Primary Decision Drivers”- Exactly one channel must be the canonical L3 source of truth, graded by the falsification gate.
- Published prose must not leak internal MIF finding identity (citation-leak gate).
Secondary Decision Drivers
Section titled “Secondary Decision Drivers”- Any exemption from the L3 gate must be declared in a manifest, never silent.
- A published format orthogonal to MIF must not be forced into L3.
Considered Options
Section titled “Considered Options”Option 1: Every channel is canonical L3
Section titled “Option 1: Every channel is canonical L3”Description: Hold blog and report identically to the full L3 conformance gate.
- Advantages: Uniform rules across every channel.
- Disadvantages: Forces internal finding identity into published prose (which the citation-leak gate exists to prevent) and over-constrains formats like pdf or audio that are orthogonal to MIF.
- Risk Assessment: technical medium; schedule low; ecosystem high.
Option 2: No canonical channel
Section titled “Option 2: No canonical channel”Description: Treat all channels as equal projections, none privileged.
- Advantages: No channel is privileged over another.
- Disadvantages: There is no single graded source of truth; the falsification bar has nothing definitive to grade.
- Risk Assessment: technical low; schedule low; ecosystem high.
Option 3: Report canonical, published channels declared-exempt
Section titled “Option 3: Report canonical, published channels declared-exempt”Description: Make the generic report channel the L3 source of truth graded by the gate; let blog and channel packs declare exemption in a manifest.
- Advantages:
reports/<topic>/<slug>.mdis the canonical L3 artifact carryingextensions.harness.verification, held to the same bar as a finding;blogdeclaresoutputs[].mifExempt: truewith a recorded reason and is kept leak-free by the citation-leak gate; genres are L3 by default (exemption is for orthogonal formats, never genres);gate_m10logs every exempt surface. - Disadvantages: Two conformance regimes (graded report vs exempt published) must be kept straight by authors.
- Risk Assessment: technical low; schedule low; ecosystem low.
Decision
Section titled “Decision”Adopt Option 3: report canonical, published channels declared-exempt. In
harness.config.json outputs[], report is the canonical L3 source of truth
graded by the gate; blog is enabled with mifExempt: true and a stated
mifExemptReason. Published channels are projections of the same artifact,
exempt only because their format is orthogonal to MIF, and only when declared in
a manifest (outputs[].mifExempt for first-class channels, pack mif.exempt for
channel packs).
Consequences
Section titled “Consequences”Positive
Section titled “Positive”- There is exactly one graded source of truth, and published prose stays leak-free.
- Exemption is auditable — every exempt surface is logged by
gate_m10, fail-closed on the report.
Negative
Section titled “Negative”- Authors must remember which channel is graded and which is a declared projection.
Neutral
Section titled “Neutral”- The blog is a first-class published projection, not a second-class output — it is simply graded differently from the canonical report.
Decision Outcome
Section titled “Decision Outcome”A canonical, gate-graded report plus manifest-declared exemptions for published formats keeps exactly one trustworthy source of truth while letting prose channels publish in their own format without leaking finding identity. The two-regime burden on authors is mitigated by deterministic logging of every exempt surface.
Related Decisions
Section titled “Related Decisions”More Information
Section titled “More Information”- Date: 2026-06-23
- Source:
docs/explanation/mif-io-conformance.md,harness.config.json(outputs[])
2026-06-23
Section titled “2026-06-23”Status: Compliant
| Finding | Files | Assessment |
|---|---|---|
| Report channel declared canonical | harness.config.json (outputs[] report) | compliant |
| Blog exemption declared with reason | harness.config.json (outputs[] blog, mifExempt, mifExemptReason) | compliant |
| L3 conformance model documented | docs/explanation/mif-io-conformance.md | compliant |
Summary: outputs[] declares the report channel canonical and the blog channel mifExempt with a stated reason; the conformance model is documented.
Action Required: None