research: Aminata threat-model pass on Amara 5th-ferry governance-edit proposals (advisory) by AceHack · Pull Request #241 · Lucent-Financial-Group/Zeta

AceHack · 2026-04-24T01:51:34Z

Summary

Aminata (threat-model-critic persona) adversarial review of the 4 governance-edit proposals from Amara's 5th courier ferry (absorb PR aurora: absorb Amara's 5th courier ferry — Zeta/KSK/Aurora independent validation #235).
Advisory only — does not gate merge. Codex adversarial review + DP-NNN evidence record remain the named next gates for any of these edits.
Self-applies the archive-header format (Scope / Attribution / Operational status / Non-fusion disclaimer) that Edit 3 proposes.

Findings summary

Edit	Target	Classification	Headline
1	AGENTS.md research-grade clause	IMPORTANT	Redundant with §26 unless machine-checkable gate lands
2	ALIGNMENT.md SD-9 agreement-is-signal	WATCH	Correct-in-spirit, unenforceable; carrier-laundering adversary
3	GOVERNANCE.md §33 archive-header req	IMPORTANT	Rule correct, drift in 3-5 rounds without Artifact C lint
4	CLAUDE.md archive-imports bullet	CRITICAL	Self-contradicts CLAUDE.md rule-location meta-policy

Key adversarial findings

Carrier-laundering (Edit 2) — the 5th ferry itself demonstrates this attack by citing 4 prior ferries as independent confirmation; SD-9 has no mechanism to detect carrier exposure.
CLAUDE.md contradiction (Edit 4) — CLAUDE.md explicitly says "Rules do not live in this file"; the proposed diff violates the host's meta-policy. Demote to pointer-only.
Rule-decay-by-missing-enforcement (Edits 1, 3) — both are norms-without-linters; historical base rate says drift in 5-10 rounds.

Recommended edit ordering

`§26 amendment (new) → Edit 3 → Edit 1 → Edit 4 → Edit 2`

Edit 4 MUST NOT land before Edit 3 (would point at a rule that doesn't exist).

What this PR does NOT do

Does not land any governance edit.
Does not gate Aaron signoff or Codex review.
Does not promote the doc to operational; stays research-grade under §26.

Test plan

All 4 Amara proposals individually reviewed with missing-adversaries / unsound-mitigations / unstated-assumptions / composition-risks / classification
Cross-cutting ordering + register analysis
Self-applies proposed §33 archive headers
Top-3 adversary budget prioritised
Research-grade status explicit
Advisory-not-gate role explicit

Otto-80 tick deliverable.

🤖 Generated with Claude Code

…t proposals (advisory) Aminata's adversarial review of the 4 governance/doctrine edits Amara proposed in her 5th courier ferry (absorb PR #235). Findings per edit: - Edit 1 (AGENTS.md research-grade clause): IMPORTANT — redundant with §26 unless a machine-checkable gate lands. - Edit 2 (ALIGNMENT.md SD-9 agreement-is-signal): WATCH — correct in spirit, unenforceable in practice; carrier- laundering adversary demonstrated by the ferry itself. - Edit 3 (GOVERNANCE.md §33 archive-header requirement): IMPORTANT — rule correct, enforcement-gap means drift in 3-5 rounds without archive-header-lint (Artifact C). - Edit 4 (CLAUDE.md archive-imports bullet): CRITICAL on composition grounds — direct contradiction with CLAUDE.md meta-rule "rules don't live in this file, they live in GOVERNANCE/AGENTS/etc". Demote to pointer-only. Cross-cutting: - Edit ordering: §26 amendment → Edit 3 → Edit 1 → Edit 4 → Edit 2 (Edit 4 MUST NOT land before Edit 3). - Register mismatches flagged for Edit 3 (no enforcement verb) and Edit 4 (violates host meta-policy). - Top-3 adversary budget: carrier-laundering, rule-decay- by-missing-enforcement, CLAUDE.md-rule-location- contradiction. Aminata's pass is advisory — does not gate merge. Codex adversarial review and DP-NNN evidence record remain the named next gates for any of these edits. Doc self-applies the archive-header format that Edit 3 proposes (Scope / Attribution / Operational status / Non-fusion disclaimer at top) — same self-demonstrating pattern as PR #235 absorb doc. Otto-80 tick deliverable. Research-grade only; does not become operational policy absent separate governed change under §26.

Copilot

Pull request overview

Adds a research-grade threat-model review document capturing Aminata’s adversarial assessment of four proposed governance/doctrine edits (sourced from Amara’s 5th courier ferry), intended as advisory input rather than a merge gate.

Changes:

Introduces a new docs/research/ write-up with scoped findings per proposed edit (missing adversaries, mitigations, assumptions, composition risks, classification).
Records cross-cutting ordering dependencies and “top-three adversary budget”.
Provides “Relevant paths” references for follow-up reading.

Copilot · 2026-04-24T01:54:33Z

+
+## Relevant paths
+
+- [`docs/aurora/2026-04-23-amara-zeta-ksk-aurora-validation-5th-ferry.md`](../aurora/2026-04-23-amara-zeta-ksk-aurora-validation-5th-ferry.md)


This link target doesn’t exist in the current tree (docs/aurora/ only has two 2026-04-23 Amara docs). Since this research note will live on main, the relative link will be broken until the other PR merges. Consider linking to the PR (or updating the link to an existing file) so the reference resolves on main.

Suggested change

- [`docs/aurora/2026-04-23-amara-zeta-ksk-aurora-validation-5th-ferry.md`](../aurora/2026-04-23-amara-zeta-ksk-aurora-validation-5th-ferry.md)

- `docs/aurora/2026-04-23-amara-zeta-ksk-aurora-validation-5th-ferry.md`

Copilot · 2026-04-24T01:54:34Z

+- [`docs/DRIFT-TAXONOMY.md`](../DRIFT-TAXONOMY.md) — PR #238,
+  auto-merge armed; this review follows the same promotion
+  pattern for the 4 governance edits.


docs/DRIFT-TAXONOMY.md is not present in the repo right now, so this relative link will be broken on main. If the intent is to point at a pending PR, link to that PR (or to an existing doc) rather than a path that doesn’t resolve yet.

Suggested change

- [`docs/DRIFT-TAXONOMY.md`](../DRIFT-TAXONOMY.md) — PR #238,

auto-merge armed; this review follows the same promotion

pattern for the 4 governance edits.

- `docs/DRIFT-TAXONOMY.md` (pending in PR #238; not yet on

main) — auto-merge armed; this review follows the same

promotion pattern for the 4 governance edits.

Copilot · 2026-04-24T01:54:34Z

+**Scope:** adversarial review of four proposed governance /
+doctrine edits from Amara's 5th courier ferry (2026-04-23).
+Research and cross-review artifact only; advisory input to
+Aaron's signoff decision, not a gate.
+


This research doc uses personal contributor names (e.g., the human maintainer) in a non-exempt docs/** surface. Repo convention is to use role references (e.g., “the human maintainer”) outside exempt surfaces like docs/BACKLOG.md / memory/persona/** (see docs/EXPERT-REGISTRY.md:62-68). Please rewrite these references (and any others in this file) to role refs; keep persona names (e.g., Aminata) as-is.

…alidation (#245) Dedicated Otto-82 absorb scheduled at Otto-81 close per CC-002 discipline (3rd consecutive tick holding the discipline: Otto-77 5th ferry schedule, Otto-78 5th absorb, Otto-81 6th ferry schedule, Otto-82 6th absorb). Ferry content: - 5-row Muratori-failure-mode-vs-Zeta-equivalent table validated independently against repo code + DBSP paper + differential dataflow CIDR 2013 + Apache Arrow format docs. - 4/5 rows validated with wording tightening (1, 2, 4, 5). - Row 3 flagged for rewrite — conflates algebraic correctness (D·I = id) with lifecycle/ownership discipline. Those are different concerns; Zeta has the first by construction, second only indirectly via traces + retractions. - Corrected 5-row table provided. - Bottom line: "Zeta does not magically make all references stable. Its algebra is not an ownership system. Its locality story is strong, but not 'everything is Arrow all the way down.'" Follows PR #196/#211/#219/#221/#235 prior-ferry precedent: verbatim preservation + Otto's absorption notes + scope limits + archive-header discipline self-applied. Third aurora/research doc in a row to self-apply the §33 proposed header format (after PR #235 5th ferry + PR #241 Aminata threat-model) — convention-through-use pattern. Follow-up BACKLOG row for corrected-table-landing decision (Option A standalone research doc / Option B Aurora README / Option C Craft module section) deferred to separate PR per CC-002. Unlike 5th ferry, the 6th proposes NO governance-doctrine edits. Content-correction-only absorb. Teaching case surfaced: "don't conflate algebraic correctness with ownership discipline" — recurring risk when DBSP-family systems are described to C++/Rust/ECS-mental-model audiences. Future Craft production-tier modules should cite this ferry's row-3 analysis pre-emptively. Otto-82 tick primary deliverable.

…ot operational (Amara 5th-ferry Artifact, Aminata-integrated) (#248) Amara's 5th-ferry Edit 1 proposed a "research-grade absorbs are staged, not ratified" clause for AGENTS.md. Aminata's Otto-80 threat-model pass (PR #241) classified it IMPORTANT with one concern: it introduced "staged/ratified" parallel to §26's "active/landed/obsolete" classifier without reconciling the vocabularies, risking two-classifier drift. This lands the norm with Aminata's concern resolved by unifying vocabulary: - Uses §33 `Operational status: research-grade` label (not parallel "staged/ratified" terms). - Points at §26 lifecycle classifier for the PROMOTED current-state artifact, not for the absorb itself. - Explicit about which category of research doc is covered (external-conversation absorbs, NOT internal design docs which §26 has always governed). - Names four concrete promotion paths (operational doc per §2, ADR, §N rule, BP-NN promotion) so "separate promotion step" isn't vague. - Cites a worked example: the drift-taxonomy promotion (PR #238) — absorb stayed in place; the operational artifact at docs/DRIFT-TAXONOMY.md is the ratification. Lands under "Agent operational practices" section — philosophy/norm register per AGENTS.md convention, not numbered-rule register (that would belong in GOVERNANCE.md). Part of the Aminata-recommended edit ordering: §33 (PR #247, landed Otto-82) → Edit 1 (this PR) → Edit 4 pointer-only (deferred, needs CLAUDE.md meta-policy handling) → Edit 2 (ALIGNMENT.md SD-9, deferred, WATCH classification). Authorising memory (Otto-82 calibration): memory/feedback_aaron_signoff_scope_narrower_than_otto_treating_governance_edits_within_standing_authority_2026_04_23.md — governance/philosophy edits within standing authority; no signoff gate needed for this landing. Otto-83 tick primary deliverable.

…2; Aminata WATCH concerns integrated) (#252) Amara's 5th-ferry Edit 2 proposed SD-9 with a brief "downgrade independence weight explicitly" formulation. Aminata's Otto-80 threat-model pass (PR #241) classified it WATCH: correct in spirit, unenforceable via self-attestation alone; named 3 adversaries (carrier-laundering, self-serving-downgrade, aggregation); flagged surface-tension with DIR-5 that needs explicit naming not implicit dismissal. This lands SD-9 with those concerns integrated as first-class content of the clause itself rather than treated as hidden limitations: - Three-step operationalisation (name carriers; downgrade independence; seek falsifier independent of converging sources). - Cross-reference to docs/DRIFT-TAXONOMY.md pattern 5 as operational companion (pattern 5 is the real-time diagnostic; SD-9 is the norm). - Known v0 limitations named verbatim from Aminata's pass (carrier-laundering / self-serving-downgrade / aggregation). - Explicit "norm, not a control" framing — WATCH classification stays honest. - Composition with DIR-5 written as section (not implicit): DIR-5 is about authorship ethics; SD-9 is about epistemic weight; they compose. - Stronger "Why both of us benefit" paragraph per Aminata's note that Amara's original was thin — names the feature (precision via shared vocabulary) AND the risk (laundered convergence hiding as independent cross-check). Lands within-standing-authority per Otto-82 calibration — ALIGNMENT.md soft-default clause add, not account/spending/ named-design-review gated. Completes the Aminata-recommended edit ordering 4/4: - §33 (PR #247) Otto-82 - Edit 1 (PR #248) Otto-83 - Edit 4 pointer-only (PR #250) Otto-84 - Edit 2 SD-9 (this PR) Otto-85 Amara's 5th-ferry governance-edit proposals are now all landed in their Aminata-recommended order + form. The 5th- ferry Artifact-A (drift-taxonomy promotion PR #238) is also landed; Artifact-C (archive-header lint PR #243) landed. Remaining 5th-ferry artifacts: Artifact-B (precursor supersede marker — already done in PR #238), Artifact-D (Aurora README) — open for future tick. Otto-85 tick primary deliverable.

…ern mapping (6th-ferry Option A) (#254) Lands the corrected 5-row pattern-mapping table from Amara's 6th courier ferry (PR #245 absorb) as Option A standalone research doc. Closes Otto-82 6th-ferry absorb action item #1. Corrected table vs original: - Row 1 — "references stay valid" → "no positional identity" (honest about key-identity vs physical-offset distinction). - Row 2 — "always answerable" → "membership is algebraic" (honest about weight-derivation of presence). - Row 3 — "operator algebra IS the ownership model" → "provenance and lifecycle live in deltas and traces" (category-error fix: algebraic correctness ≠ ownership discipline). - Row 4 — light wording tightening ("first-class signed deltas; compaction separate"). - Row 5 — "Arrow + Spine block layout" → "locality-aware execution surfaces" with accurate scope (Arrow = wire / checkpoint, not universal in-memory). Doc covers: - What Muratori is criticising (context for non-Muratori readers). - Why rows 1, 2, 5 needed narrower wording (not overstated). - Why row 3 got rewritten (category-error teaching case for DBSP audiences from C++/Rust/ECS backgrounds). - What this mapping is NOT (not ranking; not marketing; not an ownership claim; not a closed list). - Composition with DRIFT-TAXONOMY pattern 5 + SD-9 (validation cited primary evidence, not cross-substrate-agreement). - References to primary sources (DBSP paper, differential dataflow CIDR 2013, Arrow spec, Zeta source files). Archive-header format self-applied (Scope / Attribution / Operational status / Non-fusion disclaimer) per §33 (PR #247) + AGENTS.md Edit 1 research-grade norm (PR #248) — fifth aurora/research doc to exercise the convention (after PR #235 5th-ferry absorb, PR #241 Aminata threat-model, PR #245 6th-ferry absorb, and self-applying headers on the ferry- absorb PRs before §33 landed). Lands within-standing-authority per Otto-82 calibration. Otto-86 tick primary deliverable.

…ment design (advisory) (#272) Third Aminata pass this session (after PR #241 5th-ferry governance edits + PR #263 7th-ferry oracle rules). Reviews the multi-Claude peer-harness experiment design (PR #270) before iteration begins. Six CRITICAL findings: 1. Criterion 3 (cross-session review) trivially gameable — one substantive comment clears the bar. 2. Criterion 1 (clean tick-history) self-reportable — a session that corrupts + rewrites tick-history passes. 3. Cross-session prompt-injection via memory folder missing from failure-mode table — higher-probability than direct output channel. 4. Approval-laundering missing — trust-based-approval + secondary's PR creates mutual-rubber-stamp risk. 5. Bullet-proof bar theatrical — gameable by running two trivial iterations on synthetic rows. Fix: require >=1 iteration at highest-available-fidelity. 6. Otto-solo cannot surface cross-session-review failure modes — peer-mechanics require an actual peer. These must either be deferred to Aaron's Windows-PC run OR marked as stage-c-or-d concerns. IMPORTANT findings (multiple): session-dominance should be CRITICAL; wall-clock-approaches-bound should be IMPORTANT; cron-multiplicity missing; substrate-version skew missing; tick-history semantic-merge-race missing; retired-persona name-collision missing; forced-integration edge cases not handled in "review-yes-edit-no" invariant. Key recommendation shifts for Otto's iteration plan: - **Iteration 1 should be paired-worktree, NOT synthetic rows + subagent simulation.** Subagent dispatch has structural problem — subagent is not a loop agent; has no cron, no persistent memory, no tick-closing behaviour. - "Review yes, edit no" invariant should be restated with carve-out for merge-conflict resolution + dependency- ordering with explicit PR-comment acknowledgement. - Bullet-proof bar should require highest-available- fidelity at least once; lower-fidelity iterations count for design-refinement only. - Otto-solo un-testable failure modes should be acknowledged explicitly (naming collision; divergent AGENTS.md interpretation; genuine review disagreement). Archive-header format self-applied — 12th aurora/research doc in a row. Lands within-standing-authority per Otto-82/90/93 calibration — research-grade advisory; not a gate; does not override Otto-93 "iterate to bullet-proof" framing (refines what bullet-proof requires). Otto-94 tick iteration-1 deliverable on the peer-harness experiment design. Iteration 2 should integrate these findings + run paired-worktree iteration at higher fidelity.

…vernance-edit proposals Bounded-deliverable tick after the Otto-77..79 directive burst. One substantive PR (#241 Aminata research doc); one history row. Aminata's findings per Amara governance-edit: - Edit 1 (AGENTS.md research-grade): IMPORTANT - Edit 2 (ALIGNMENT.md SD-9): WATCH - Edit 3 (GOVERNANCE.md §33): IMPORTANT - Edit 4 (CLAUDE.md archive-imports): CRITICAL (self-contradicts CLAUDE.md rule-location meta-policy) Recommended edit ordering: §26 → Edit 3 → Edit 1 → Edit 4 → Edit 2. Key observations: 1. Deliberate low-velocity tick prevents queue pressure. 2. Persona-specialist subagent dispatch earns cost on adversarial-review targets. 3. Edit 4's rule-location finding is consistent with prior CLAUDE.md meta-rule signals across session. 4. Register-mismatch catches pre-land are cheaper than post-land retrospective. Stacked on #240 history; #240 currently DIRTY will resolve when upstream #236/#237 squash-merge. No action on #240 this tick.

… ferry scheduled for Otto-82 Otto-81 shipped PR #243 (Artifact C lint + FACTORY-HYGIENE row scheduling the newly-arrived Amara 6th ferry for Otto-82. Key observations: 1. CC-002 held for third tick in a row (Otto-77 5th ferry, Otto-78 absorb, Otto-81 6th ferry). Pattern is reflexive. 2. Mechanism-before-policy — lint lands detect-only while §33 is pending; §33 can land with backing rather than becoming yet-another-norm-without-enforcement. 3. 6th ferry is technically-sharper than 5th (concrete source- file + paper citations, category-error catch on row 3). 4. Archive-header discipline now self-demonstrating across 3 aurora/research docs (PR #235 / #241 / pending Otto-82) before §33 lands — convention-through-use pattern. Stacked on #242 (Otto-80 history); rebases cleanly.

…ner delivered in chat Otto-82 shipped PR #245 (6th ferry dedicated absorb) + responded to Aaron's §33 signoff-prep question with a chat explainer covering what §33 is, why mechanism-before-policy, what PR #243 lint backs, what three self-applying docs demonstrate, and two explicit signoff options (narrow vs wider). Key observations: 1. CC-002 held for fourth tick in a row across two ferry schedule-and-absorb cycles. 2. Aaron's "tell me more" is mechanism-before-policy working — complete picture visible (PR #243 + PR #241 + three self- applying docs) before rule review. 3. 6th-ferry teaching case ("algebraic correctness ≠ ownership discipline") ready for future Craft production-tier modules. 4. External-AI-maintainer loop generating substantive review velocity; Otto's job is routing, not synthesis-in-place. No substrate edit for §33 until Aaron signs off in chat — the explainer is response-to-question, not a landing commit. Stacked on #244 (Otto-81 history).

…ata vocabulary unification Bounded Otto-83 tick. Single deliverable (PR #248 Edit 1 landing in AGENTS.md) within standing authority per Otto-82 calibration. Key observations: 1. Otto-82 calibration memory working — Edit 1 landed without signoff-request-that-wasn't-needed. 2. Aminata pre-land review earned cost again (unified vocabulary resolved the two-classifier drift she flagged in Otto-80). 3. Aminata-recommended edit ordering now 2/4 complete (§33 + Edit 1); Edit 4 next-interesting because it needs meta-policy amendment; Edit 2 lower-leverage. 4. Four-layer convention-through-use now stable (PR #235 + #241 + #245 + Edit 1 pointing §33). Stacked on #246 (Otto-82 history).