**Status.** Provisional 12-part SMALL-band sample, not an approved content budget, starting inventory, progression, balance model or implementation. The owner has selected connected reusable parts, shared oxygen/equipment/loot capacity, carried-property return on death, mostly automatic ordinary fire and manual specials. The exact port grammar, rig growth, acquisition and every number below remain proposals.

The worn emergency cutter and suit functions are not catalogue parts: they have no ports and do not provide a second automatic weapon. Ten entries preserve the eight illustrated Build Rules parts plus the selected opening’s Threat Sensor and Manual Trigger. Stride Switch and Vector Nozzle fill movement-event and spatial-control gaps without creating a stat tier.

A **packet** contains a trigger and may carry charge, heat, force or fragments. A source creates a new external wave only when its rule says so. Descendants keep that wave; each part resolves at most once in it. Proposed cost is written **C/A**: overall carrying units / active rig spaces. An installed part pays its carrying cost once, not again as cargo. An inactive carried part pays C but no A. The opening parts use 1/1 so the current four-space reconfiguration remains legal; larger examples state their separate later-area assumption.

| **Fork Relay** | One packet input, visibly ordered outputs 1/2. It copies trigger, payload **type** and wave identity. A bare trigger reaches every accepting branch. Usable whole U divides evenly between two pre-allocation candidates, output 1 taking an odd remainder; a sole candidate gets the full payload. | One service-board pulse visibly reaches two lamps, with invalid sockets refusing before commitment. | **1/1.** Link/type/readiness/storage/target-blocked branches are excluded before allocation. A post-allocation sub-threshold share refuses and dissipates rather than being copied, buffered or reassigned. Packed Fork gives no benefit. | Sensor → Fork → Projector + Barrier uses a bare trigger; charged release → Fork → Barrier + Arc divides usable charge between both candidate receivers. | Loose in Cargo Turn’s maintenance cradle; may serve infrastructure or rig. |

| **Capacitor** | Charged-packet input, exact **0–4 U** stored state, chosen release trigger and one charged output. Manual **2 U** release starts a new wave; release caused by a rig packet retains that packet’s wave. Becoming ready/full emits nothing. | Rising meter, visible saturation and a controlled discharge into a dead load. | **1/1.** Accepts at least 1 U up to free space and visibly vents delivered overflow; at 4 U it is blocked before Fork allocation. A release below 2 U or toward a refusing first receiver does not commit and retains charge. | Dynamo → Capacitor for horde-fed storage; manual release → Lens → Projector for a burst. | Defibrillator bank, door buffer or damaged core whose stored state stays visible. |

A part may fit a visibly compatible service port: Fork can split a service pulse, Dynamo can expose an impact response, and Nozzle can test a pressure path. Installation uses world time and leaves the part outside carried property; completed work and installed state follow their owning persistence rules. Individual behavior and direct-port validity may be shown. A part name or port never translates alien text, identifies an unseen destination, proves a historical interpretation or chooses the whole-station binding mechanism.

**First-pass package planning envelope:** Systems authoring **10–16 role-hours**, one independent review **3–5**, and one bounded response **2–4**; **15–25 role-hours total**. The author range maps inside the Whole Game Outline’s 10–18-hour author-only catalogue allowance; review/response are additional and must not be counted again as that allowance. These are planning estimates, not logged work or throughput.

**Later detailed SMALL catalogue specification:** Systems authoring **36–60 role-hours**, review **8–12**, response **4–6**; **48–78 role-hours total**. Unique work would define per-part port diagrams, state/overflow variants, source-placement alternatives, legal-link matrix and authored edge cases. It excludes UI-owned cue/accessibility presentation, acquisition economy, implementation, assets, numerical tuning and playtesting. TARGET/STRETCH expansion remains unestimated until reuse and per-part authoring cost are known.

Later owner-authorised implementation and real-player observation must establish whether individual parts are legible, the four-space swap is discovered, combinations remain expressive without one dominant path, costs compete meaningfully with oxygen and loot, field edits are usable under pressure, and rare station uses feel revelatory rather than like metadata keys. None is established here.