sovereign-agent

Ruling: Unit 11 scope — pilot start marker, multi-SKU catalog, Chapters 8-12

• id: ruling-2026-08-30-unit11-scope
• decided: 2026-08-30
• authority: principal
• applies_to: Sovereign Agent 1.x, Unit 11 — scope, not implementation
• status: ACTIVE

This ruling records the Principal’s decision on the six open questions a read-only Unit 11 scope descent surfaced but explicitly did not resolve. The descent itself (.unit11/SCOPE-DESCENT.md at the time of this ruling, untracked investigation evidence, corrected additively after Sparring’s independent review found a false namespace-mention claim, not a durable record) is superseded as the citable scope authority by this file. No code, curriculum content, or SOW exists for Unit 11 yet; this ruling authorizes scope only, not implementation, and not the SOW’s own authoring — see “Next authorized action” below.

The six questions this ruling answers

1. The pilot start/finish seam. Amendment 6 (docs/sows/sovereign-agent-v1-educational-control-plane.md:62-68) splits one 30-day pilot across two units — Unit 11 “starts” it, Unit 12 “finishes” it — with no operational definition anywhere in the ratified record of what “starting” concretely produces.
2. The Store-expansion dimension. Amendments 1 and 6 both say Unit 11 “expands” the Store’s single-SKU walking skeleton, without naming along which dimension.
3. Chapter-to-feature mapping for Chapters 8-12. No ratified document names any chapter title, count, or subject for the five new chapters.
4. The Unit 12 soak’s own definition. Squarely Unit 12’s territory, but a Unit 11 SOW needs enough forward context to avoid building features awkward for a not-yet-defined evaluation process.
5. The docs/v0.6-soak.md naming collision with the unrelated Unit-12-owned “Andrea-profile soak” concept.
6. Whether an analogous authority-language tension exists for Unit 11’s own “expands”/”lands”/”starts” verbs, matching the genuine tension Unit 10’s own descent found between amendment 6’s “completes” and docs/units-0-6-contract.md’s “expands, reorganises and polishes” language for Chapters 0-7.

Holding 1 — the pilot start/finish seam: a bounded operational marker plus a governance receipt

Adopt a bounded form of option (d) named in the descent: an operational marker plus a governance receipt, not a hollow prose claim and not a mechanism that reaches into Unit 12’s own finish/verify/soak/release territory.

Unit 11 must implement an explicit, durable pilot-start act producing:

• A first-class, queryable SQLite record — not an inference from prose or arbitrary JSON, matching the same discipline the governing Unit 9 ruling already required for Pulse attribution (docs/rulings/2026-08-29-unit9-pulse-is-separate-from-supervisor.md).
• An append-only pilot.started event or equivalently explicit event kind.
• A stable pilot identity.
• A canonical UTC start time.
• The Store organization and pilot-profile identity.
• A reference to the evidence namespace where Unit 12 can later assemble the redacted proof pack (docs/sows/sovereign-agent-v1-educational-control-plane.md:21’s own done_when phrase — “30-day Sovereign Store pilot -> redacted proof pack accepted” — the sole ratified source naming what completion produces).
• Idempotent behavior: replaying the same start request cannot create another pilot.
• A refusal when an incompatible pilot is already active.

Unit 11 must NOT implement pilot completion, elapsed-time acceptance, soak scoring, credentialed verification, proof-pack acceptance, or release — those remain entirely Unit 12’s.

The operational marker starts the clock. A short append-only governance receipt must cite the real marker without pretending the pilot has finished.

The pilot does not start merely because Unit 11 code merges. After implementation acceptance and clean-main verification, the Principal must separately authorize Master to execute the pilot-start act against the named pilot organization. Unit 11 closes only after that act and its governance receipt are durably verified — a fourth authorization gate beyond the SOW-review, implementation-review, and acceptance gates every prior unit has already used.

Unit 12 may later consume this marker, but it cannot claim the 30-day pilot complete before 30 actual days have elapsed from the recorded timestamp.

Holding 2 — the Store-expansion dimension: a small multi-product catalog

Adopt a strengthened form of option (a) named in the descent: expand the walking skeleton from one hardcoded product (Product(sku="SKU-TEA", ...), src/reference_organizations/store/__init__.py:60) into a small multi-product catalog.

Minimum accepted behavior:

• At least two distinct SKUs.
• Independent stock levels and reorder points.
• Independent supplier/replenishment routing where the existing model permits it.
• A sale or inventory change for one SKU cannot wake or replenish another.
• Multiple qualifying SKUs can each create one canonical governed replenishment chain.
• Replay, restart, and concurrency preserve per-SKU idempotency.
• The existing inventory.changed -> wake gate -> Pulse -> replenishment vocabulary remains intact.

Unit 11 must NOT add new signal kinds, effect kinds, governance concepts, or core organizational primitives merely to make the expansion appear larger. The point is proving the accepted pipeline generalizes beyond its single-SKU teaching fixture, not inventing new production surfaces.

If implementation genuinely requires more than the remaining source-budget headroom (27/40 modules, 6139/6250 nonblank lines, 7/30 root exports at this ruling’s base commit — 111 lines of headroom, confirmed by scripts/verify_source_budget.py), the SOW must request a precise budget amendment before code is written, matching F-U9-1’s own precedent. Correctness must not be compressed to fit 111 lines, and a larger ceiling must not become a target.

Holding 3 — Chapters 8-12, the exact chapter map

Chapter	Directory	Teaching responsibility
8	ch08_the_store_becomes_a_catalog	Replace the single-product fixture with a genuine multi-product catalog.
9	ch09_each_product_has_its_own_threshold	Independent stock state and reorder decisions per SKU.
10	ch10_one_signal_wakes_one_need	Signal-to-SKU binding; no cross-product wake or replenishment.
11	ch11_replenishment_scales_without_losing_governance	Multiple governed replenishment chains while preserving idempotency, evidence, and fencing.
12	ch12_the_pilot_begins_with_a_receipt	Execute and inspect the pilot-start mechanism while distinguishing “started” from “finished.”

Each chapter must:

• Import and execute production behavior.
• Include the instructor-note structure established by Unit 10 (book/CONTENT-SOURCE.md:48-58, seven required sections per INSTRUCTOR.md).
• Participate in the existing mechanical chapter, sequence, link, and frontmatter checks (scripts/verify_curriculum.py, extended by Unit 10).
• Avoid teaching forks or fabricated pilot/Pulse evidence.

Chapter 12 teaches the pilot-start mechanism in an isolated exercise. It must not accidentally start the real 30-day pilot during the curriculum gate — the chapter’s own exercise must run against an exercise-scoped, disposable pilot identity, never the real named pilot organization the separately-authorized pilot-start act (Holding 1) targets.

Holding 4 — Unit 12’s own soak definition is not decided here

No Unit 11 decision is made about duration, sampling, participant count, repeated sessions, aggregation, or pass/fail criteria for the Andrea live evaluation.

Unit 11 must only preserve future evaluability:

• deterministic offline exercises;
• durable pilot identity and start time;
• structured evidence references;
• no dependence on unavailable credentials.

Unit 12’s own descent and SOW must define the human evaluation and soak protocol.

Holding 5 — soak terminology, a naming rule for future work

Historical v0.6 documents (docs/v0.6-soak.md, site/v0.6-soak/) remain unchanged.

Correction, additive, matching the descent’s own already-corrected citation: docs/v0.6-soak.md itself references scripts/soak_single_node.py at its own lines 7 and 10 — that script does not exist in the current tree; this is pre-existing v0.6-era documentation rot, not something this ruling or the descent that preceded it introduced or is required to fix. The naming-collision concern this holding addresses stands entirely on docs/v0.6-soak.md’s real “72-hour infrastructure soak” content, independent of that stale script reference.

Future documents must avoid the unqualified word “soak” when ambiguity is possible:

• Use 30-day Store pilot for the operational production run.
• Use Andrea live evaluation for the timed human learner session.
• Use v0.6 infrastructure soak when referring to the historical 72-hour infrastructure exercise.

This is a terminology rule for future work, not a Unit 11 implementation feature.

Holding 6 — no analogous authority-language conflict exists; closed

Amendment 6 governs Unit 11. docs/units-0-6-contract.md contains zero mentions of Unit 11; the four Unit 12 mentions it does contain (docs/units-0-6-contract.md:92,119,135,201) all defer credentialed provider smokes and do not describe Unit 11’s mission. No corrective ruling or historical annotation is required for that document.

Correction, additive: an earlier draft of the descent this ruling supersedes claimed docs/units-0-6-contract.md “contains zero mentions of Unit 11 or Unit 12” — false on the Unit 12 half, caught by Sparring’s independent re-verification before this ruling was drafted. The document’s silence specifically on Unit 11’s mission — the fact this holding actually depends on — was accurate throughout and is unaffected; only the overbroad “zero Unit 11/12” phrasing was wrong, corrected here rather than silently carried forward.

Next authorized action

Master may file this decision as this append-only Unit 11 scope ruling and update the rulings index. This filing must:

• Transcribe these holdings faithfully.
• Preserve the corrected soak and namespace citations.
• Contain no implementation SOW or production/curriculum changes.
• Pass the documentation and curriculum gates.
• Receive exact-head independent Sparring review before merge.

After the ruling is merged and verified on clean main, Master may prepare — but not implement — the Unit 11 SOW. The SOW must encode the pilot-marker transaction and idempotency proofs, the multi-SKU isolation matrix, the five-chapter mapping, the real-pilot authorization boundary (Holding 1’s separate Principal authorization gate), and any necessary source-budget request.

No Unit 11 implementation, real pilot start, credentialed provider smoke, Unit 12 evaluation, proof-pack acceptance, or release work is authorized by this ruling.

How to check this ruling against the repository

# Confirms the current budget headroom this ruling's Holding 2 cites
python scripts/verify_source_budget.py

# Confirms the exact Unit 12 mentions in units-0-6-contract.md and their content
# (deferring credentialed smokes, never describing Unit 11's mission) --
# Holding 6's own corrected claim
grep -n "Unit 11\|Unit 12" docs/units-0-6-contract.md

# Confirms the single-SKU starting point Holding 2 requires expanding
grep -n 'Product(sku="SKU-TEA"' src/reference_organizations/store/__init__.py

# Confirms no pilot mechanism exists yet -- Holding 1 requires Unit 11 to build it
grep -rni "pilot" src/ scripts/ || true   # must be empty until Unit 11 lands

# Confirms the v0.6 naming-collision content Holding 5 addresses, and its own
# pre-existing stale script reference (not introduced by this ruling)
grep -n "soak_single_node" docs/v0.6-soak.md

# The rulings index and directory agree
python scripts/verify_curriculum.py

This site is open source. Improve this page.