# IF Process Changelog (Full Current)

Source (canonical): `docs/703-if-whitepapers-bible-v4.19-2026-02-25T113459Z.md`

Patches:
- v2.2: diagrams requirement + lint preflight; clarified anchor line + italic stress-test (no meta labels).
- v2.3: lint-fail if visible template labels appear in publishable docs.
- v2.4: restored high-impact narrative cadence (headline, boardroom decision line, contrarian checkpoint, psychology closer, optional dry humor) while keeping output free of template-marker labels and personal names.
- v2.5: adds reusable decision narrative profile (seven-part executive arc) and external reviewer packet (boardroom + ML + security + execution) to reduce rework per whitepaper.
- v2.7: adds front-door problem statement and goal, codifies LLM-time execution windows (30/60/90 minutes and 3/6/9 hours), operator real-time checks, day-scale user-testing reality, and requires this front matter in output documents.
- v2.8: requires an explicit top-of-document `Who|Why|What|Where|When|How` block and applies that requirement to output documents.
- v2.9: adds baseline-timing conflict handling for multi-lane pilots, requires schema-defined source-of-truth config artifacts, adds KPI phase-applicability rules for calibration windows, and standardizes applied-plan status plus dated follow-up task sections for revisioned ops papers.
- v3.0: clarifies Layer 2 (authoritative bindings) vs Layer 5 (discovery catalog), adds cross-document reference protocol, adds supersession/retirement and complexity-budget rules, and scopes Section 13.1 enforceability language to operational sections.
- v3.1: keeps contrarian framing and adds an explicit “assumption most likely to be wrong” lens, self-applies complexity-budget justification, adds a stack diagram, tightens naming and claim semantics, adds enforcement + maintenance discipline, and hardens schema stubs.
- v3.2: self-applies status/review metadata and LLM-assist disclosure, clarifies link conventions and enforcement equivalence, labels diagram relationships, adds hard-ref lint spec hook, adds link-liveness cadence checks, and documents schema-intent for open sub-objects.
- v3.3: codifies top-of-document audience navigation as mandatory for decision-facing papers and adds lint-enforceable audience-nav checks.
- v3.5: integrates all Section 13 lenses into the primary writing contract, mandates single-voice delivery (no visible scaffold labels like Punch/Reframe/Psychology), and keeps the four-audience explainer as a non-negotiable top-level requirement.
- v3.6: codifies paragraph-first reporting, sharp anchor-line quality, explicit non-linear section-numbering disclosure for audience-priority layouts, consistency checks between required fields and sample schemas, and front-matter approver/disclosure requirements from the latest whitepaper cycle.
- v3.7: adds pack-level append-only discipline for multi-audience distribution bundles (new pack per revision), forbids in-place patching of previously numbered packs, and standardizes bootstrap/fail-closed behavior for future pack updates.
- v3.8: hardens outward path policy: publishable docs must never include absolute local-root literals and must render local/internal references only as `{$path}/...`.
- v3.9: moves style-guide metadata from header to footer (`Style Guide: Whitepaper vX.Y`), requires `Problem statement` immediately after `Who|Why|What|Where|When|How`, and replaces `{$path}` token dependence with dual path policy (repo-relative internal paths + public no-login URLs).
- v4.0: codifies cross-module review hardening from `if.bus` revisions: required acronym expansion on first mention, mandatory evidence hierarchy with independent-vs-operator-assisted boundaries, required `can conclude / cannot conclude` reviewer block, mandatory minimal external verification command set (3-5 commands) separate from full operator set, explicit promotion path/date for operator-assisted evidence, continuity requirement beyond single-owner dependency, and footer version clarity (`Style Guide` profile vs `if.whitepapers.bible` version).
- v4.1: codifies depth hardening from `if.api` and bible review cycle: requires concrete implementation examples (spec + normalized output) for module explainers, requires adapter-maturity classification when inventory breadth is cited, standardizes evidence-hierarchy column set and cross-module-link rationale, clarifies operator-only absolute-path policy, adds CI status/reminder discipline, adds explicit current-enforcement + gate-failure behavior for Section 12, and promotes reusable URL-liveness tooling + index-collision preflight guidance.
- v4.2: codifies `if.story` editorial hardening: requires explicit monitoring-boundary block for preview modules, requires escalation wording template when claim strength is uncertain, enforces `.md.txt/.json.txt` consistency in minimal external verification commands, standardizes concrete promotion artifact shape for operator-assisted evidence, and tightens first-mention definition rules for core conceptual metaphors/pairs.
- v4.3: minor restructure and clarity hardening: adds a scan-first checklist at Section 5.3, makes Section 12 gate status/failure behavior explicit, and aligns link-liveness tooling status to current repo reality (`proposed` until script exists).
- v4.4: records provenance of boardroom-explainer paragraph style: originated in the boardroom layer and normalized into Section 5.2 decision narrative profile.
- v4.5: promotes explicit contrarian alternate-view and psychology-friction checks into scan-first checklist and decision narrative additive lens.
- v4.6: rebuilds Section C to reflect actual stack state (`canonical | integrated | exists | planned`) and adds explicit visuals-rules plane status.
- v4.7: codifies control-cutover interpretation discipline for coverage snapshots: when controls (for example signatures) are enabled on historical append-only stores, require explicit cutover timestamp, pre/post-cutover counts, and a trajectory statement so early percentages are not misread as implementation neglect.
- v4.8: codifies external communication integrity hardening from multi-model audit feedback: requires audience-nav purpose phrasing, explicit operator-vs-auditor section boundary, iterative diagram fail-loop annotation (`patch + regenerate`), lens operational definitions + conflict/escalation rule + independence disclosure, deterministic gate scope line (artifact-only), source hierarchy split (`language standards` vs `implementation pattern references`) with rule-to-source traceability map, human-readable as-of timestamp plus source boundary for time-sensitive metrics, PII handling boundary in security sections, explicit progression criteria when RC/baseline decisions are used, and replay-placeholder appendix discipline (`{$path}` only in clearly labeled internal appendix with generic descriptors).
- v4.9: adds architecture-paper hardening from external reviewer cycle: explicit spec-spine requirements, attacker-model and control-plane failure-mode requirements, approval-fatigue/human-factor checks, and confused-deputy + replay-chain reviewer challenge requirements.
- v4.10: hardens path discipline and source traceability for future docs: local-path lint must ignore external URL path segments, publish checks must use `scripts/lint_if_whitepaper_paths.py` (no forced `{$path}` placeholders unless strict-masking mode is explicitly enabled), and architecture papers must include source access timestamps plus repo tag/commit identifiers where feasible.
- v4.11: codifies second-round trust-contract hardening for architecture papers: requires canonicalization + approval-binding surfaces (`verb/tool_id/adapter_version/policy_digest/normalized destination/evidence digest`), deny-code taxonomy, forensic replay bundle manifest, immutability tier split for latency claims (Tier A local chain vs Tier B external anchor), explicit validation-window and unauthorized-action definitions in go/no-go gates, and secret-redaction test contracts (including canary leakage checks).
- v4.12: codifies release-final hygiene from third reviewer round: requires version-aligned feedback-closure matrix and changelog entries, requires redirect-chain treatment as observed receipt evidence (with high-risk pre-commit interception), requires control-plane isolation abuse test (`AT-06` equivalent), requires explicit SLO-scope line for latency targets (what is included/excluded), and requires redaction-proxy latency target when secret-isolation controls are claimed.
- v4.13: codifies final trust-accuracy hardening: requires OWASP class references to use canonical IDs (`LLM01`, `LLM02`, `LLM05`) when cited, requires deny-taxonomy completeness versus abuse-table expected codes, requires mandatory evidence-digest binding for high-risk approvals (`send/auth/payment/admin`), requires policy-update governance language (CI/CD + multi-party sign-off), requires semantic-target execution validation (visible/unobscured/in-viewport), and requires split Tier-B anchoring targets for critical receipts vs evidence payloads when external anchoring is claimed.
- v4.14: codifies architecture-closure discipline for future papers: requires post-`3/6/9h` timeline specificity (Day 3/Day 14/Day 15+ or equivalent), requires explicit audience overlap disambiguation in audience-nav rows, requires expanded trust glossary for all trust-critical terms, requires explicit model-to-envelope translation boundary language, requires constructor-integrity language (canonical hash recompute + live target recheck before execution), requires signed policy-promotion-receipt enforcement for active policy digests, requires staged validation coverage criteria + named regression-suite scope + at least one composed/cascading abuse case, requires SLO provenance wording to mark numeric targets as internal calibration (`Inference`) rather than external-standard mandates, requires reviewer packet question for graceful degradation under partial control-plane failure, and requires appendix parity fields (`environment_parity_attestation`, `compensating_controls`, `regression_baseline`) when release gates depend on staged evidence.
- v4.15: codifies implementation-practicality and operations-clarity hardening for future architecture papers: requires cross-team coordination cadence notes for hardening windows, requires a concise audience quick-reference block plus glossary-to-section cross-references, requires at least one anonymized failure-pattern example in high-stakes problem statements, requires attacker-model mitigation-priority annotation, requires a new-tool evaluation checklist when ecosystem matrices are presented, requires at least one implementer acceleration artifact (schema-validation references, canonicalization pseudo-code, or CI/CD gate snippet), requires explicit contingency rules for failed execution milestones and false-positive denial handling, requires sample observability rules or query patterns when burn-rate/SLO claims are made, requires reviewer-packet sample output shape, requires feedback-closure impact column, requires standards-update monitoring guidance, and requires an appendix section for deviations-from-production when release evidence depends on staging environments.
- v4.16: codifies release-polish discipline from v1.8 review pass: requires a quick table-of-contents block for long architecture papers, requires explicit bootstrap semantics for first-action lineage fields when session chaining is introduced, requires a provenance/rationale note whenever fixed staging floors (for example `N >= 200`) are used, requires mitigation-priority scales to be at least three-level or explicitly justified as binary, and requires glossary coverage for Tier A/Tier B immutability whenever those terms appear in go/no-go or ops sections.
- v4.17: codifies debate-hardening discipline for trust-control papers: requires explicit approval-artifact signer metadata when `required_approvals` are used, requires tool-capability mismatch enforcement (`verb` vs tool map) plus deny taxonomy alignment, requires Tier-B anchor evidence fields in receipt schemas and gate templates (`anchor_reference`/`anchor_timestamp` or equivalent), requires policy-promotion gate checks for step-up signer attestation before digest activation, and requires composed-attack test telemetry fields (turn ids, redirect hops, approval receipt hashes, timestamps) when narrative-chain abuse tests are claimed.
- v4.18: codifies final consistency discipline for release-candidate trust papers: requires determinism claims to be scoped to pinned runtime/input tolerances (not open-web absolutes), requires policy-rule to deny-taxonomy parity (every enforced rule has an explicit deny code), requires explicit evidence-payload parent linkage fields in receipt schemas (`parent_chain_index` or `parent_receipt_hash`), requires high-risk verb approval rules to enumerate `send/auth/payment/admin` explicitly, requires a defined false-denial denominator (`legitimate high-risk attempt`) when false-denial SLOs are claimed, and requires standards as-of timestamps to align with document snapshot metadata.
- v4.19: codifies anti-meta writing discipline for management and external papers: body sections must not discuss "which bible/skill/process was used" unless the document's objective is documentation governance; normal papers should keep only the footer `Style Guide: Whitepaper vX.Y`, with optional `Writing Standard Source` only when strict disambiguation is required.

