# S006: use a named retained result, read its original source and return

**Four bounded existing-actor paths Exercised · underway and unjudged · independent qualification pending**

[Current progress](/development/progress.html) · [Work and evidence](/development/progress-scope.html) · [Exact start and prior history](/development/progress-history.html#s006-start)

The named-result journey now returns the full original source through the installed semantic host and Content owner. Exact input and actual owner result survive both environments reopening. Useful staged CLI/MCP output, operator source/Back and an assisted fresh CLI sample are observed. Negative/refusal/count packets have returned; **Aspen’s qualification judgment remains pending**. Rowan leads; Juniper supports integration/reporting. Main S006 remains underway and unjudged.

This plan and progress share fixed native cutoff **4976**, recorded **8 October 2026 at 10:56:59 UTC / 04:56 MDT**. S005 finite guidance remains Qualified/completed, S004 remains not completed and all sixteen full Version 1 outcomes remain incomplete. Later judgments are excluded. The exact source09404cb5/cutoff4900 start remains history.

---

## Selected result and maturity

The selected outcome is unchanged: use a meaningful named retained result to read its exact original source through actual semantic assessment, receive useful ordinary output, deliberately inspect unchanged references and reopen the same result/action. Actual owner permissions remain authoritative; a separately labeled current observation never substitutes for the captured original.

| Selected boundary | Maturity at cutoff | Actual bounded evidence and limit |
| --- | --- | --- |
| Named retained result and reopen | **Exercised** | Actual full result/input/owner observation survived both environments reopening without reseeding. Existing actor, not fresh enrollment; independent qualification pending. |
| Semantic original-source action | **Exercised** | Installed host assessed the full alternative input contract, selected the exact-content branch and returned all 876 original source characters equal to an independent exact read. Negative/refusal packets are returned for judgment. |
| Ordinary CLI/MCP/Lisp output and deliberate inspection | **Exercised** | Staged CLI and temporary MCP public default expose useful complete text; deliberate inspection preserves exact input/result/report. Fresh CLI is assisted; independent unfamiliar MCP and connected harness deployment are not established. |
| Operator shared action, full viewport, Back and reopen | **Exercised** | Actual matched operator displayed the entire short source; Back returned to the same named result/observation. All 876 characters fit at 80×24; no long-note scrolling or independent GUI acceptance follows. |

Exercised means a bounded ordinary journey with relevant recovery was demonstrated. Readiness, source tests, publication/staging, installation, runtime and independent qualification are distinct. This is a progress snapshot, not a closeout or blanket Qualified claim.

## Public source and actual availability

| Responsibility | Published source | Delivery / use |
| --- | --- | --- |
| CLI/MCP presentation | [c2f0ab27](https://gitlab.com/sidecat-dev/sidecat/-/commit/c2f0ab270701ba297f7a671efa0fa192b127351e) | Matched CLI/maintain candidates staged; actual CLI and temporary staged MCP use observed. |
| Operator UI | [6b567afa](https://gitlab.com/sidecat-dev/sidecatui/-/commit/6b567afa01b5dc46b9ab6980579abfb22cadea21) | Matched console/gateway staged; actual original-source viewport and Back observed. |
| Shape Engine | [6460491f](https://gitlab.com/sidecat-dev/sidecat-shape-engine/-/commit/6460491f7380abd606ec2c1e74f20f77f75edb88) | Branch-aware declaration/schema closure source published and consumed by the shared host. |
| Semantic assessment host | [23676ba1](https://gitlab.com/sidecat-dev/sidecat/-/commit/23676ba1ecba3f1ab86fe424f46b30b3075b74fe) | Focused source run passed 8 top-level / 12 leaf cases; built, installed and active on the existing Team node. |
| Retained-result helper | [a9110345](https://gitlab.com/sidecat-dev/sidecat-packages/-/commit/a9110345fbb47e0ea61608d777abee07b871593d) | Published; actual use is legitimate native kept adoption, **not maintained-package installation**. |

The read-only `semantic-assessment.assessment.check` query assesses supplied operation, whole schema closure and captured request; it does not fetch Content or execute effects. The helper associates `/fields/source` with `content` and dispatches `content.note.read` only from the complete ready exact-original preview. Exact carried references remain behind deliberate inspection, with authored text unchanged.

Actual Content output includes the full 876-character original, title and six-field reference. Complete owner result/input equal the independent exact read. A first order-sensitive JSON-string comparison was corrected to canonical field comparison without changing results. Both environments reopened without reseeding and preserved the whole observation, including actual owner result rather than a reconstructed preview.

## Actual interface use and assisted fresh trial

The matched operator opened the full source viewport and returned to the same named result/retained observation, exit 0. Two opening keys overlapped asynchronous menu loading; an additional explicit open displayed the source. The complete text fit at 80×24, so Home/End produced no movement. This does not qualify long-note scrolling.

A temporary staged MCP server returned the same complete source. Public default output contains state/title/text/observation/actions; explicit JSON and transport inspection preserve exact owner input/result and report. This is real-data transport evidence, not an independent unfamiliar MCP trial or an update to the running harness connector.

A fresh **Sol/high** worker completed CLI source/inspection/reopen/return, with initial and returning output byte-identical. One parent clarification mapped the actor ID to its friendly name; the participant independently checked scope, without a profile switch. Aspen compared the saved receipts without rerunning the helper. This scoped receipt review is separate from the pending overall qualification judgment.

The participant reported **18 shell submissions / 13 Sidecat commands / 5 helper calls**: nine startup/discovery, three initial task use, **three returning submissions** and three evidence setup/writes; zero failed shell submissions or retries. Returning use is not a two-call result. The first precise timestamp is missing and first saved stdout was reconstructed from immediately read bootstrap files; later outputs are actual tool returns. Nested host/Content calls were not measured by this participant. Fresh worker context used an existing actor, not fresh enrollment or human identity.

## Separate measured business calls

A delegating tracer invoked the original native callable and retained its actual response; it manufactured no assessment or owner result. These measurements are separate from the fresh participant’s 18-submission trial.

| Actual stage | Assessments | Content calls | Scope |
| --- | ---: | ---: | --- |
| Name-only menu | 0 | 0 | No business dispatch. |
| Explicit original-source action | 1 | 1 | Full source and exact actual owner result. |
| Carried inspection | 0 | 0 | Reads the retained observation. |
| Inspection after both environments reopen | 0 | 0 | Same canonical carried observation; no reseed. |
| Returning explicit source action | 1 | 1 | A deliberate new read, distinct from inspection. |
| Each of four pre-dispatch negative actions | 1 | 0 | Assessment non-success; Content not dispatched. |
| Genuine native owner-refusal action | 1 | 1 | Exactly one owning read attempt; no retry/fallback. |

The source/count trace captured 29 CLI submissions: 26 exit 0 and **three preparation/control failures**. It separately reports five helper calls, seven helper-observation calls, 22 measured logger calls plus three control logger calls, 11 log reads and one normal preparatory Content read. Logger/observation overhead is not user/business call count. Individual logger reply bodies/keys are absent, so this is not a daemon-audit claim. Setup/control failures remain failures, not product passes.

## Negative and real refusal evidence

Four actual helper actions each made one assessment and zero Content calls, returning public aggregate incomplete/action unavailable:

- Missing source member: found false, blocked, incomplete preview.
- Invalid digest against the **unchanged captured live schema**: incompatible exact-pattern cause.
- Unsupported `unevaluatedProperties`: unknown with reason preserved, using a **changed synthetic declaration supplement**.
- Overlapping `oneOf`: ambiguous input contract with both branches compatible, using a **changed synthetic declaration supplement**.

Supplied synthetic values/declarations are labeled; the installed host and schema were unchanged. No mock assessment report supplies these results.

A separate raw installed-host competing-value control returned two compatible suggested connections, selected neither, with empty arguments/incomplete preview and blocked unresolved Content. It did not produce a reachable two-ready-candidate helper case or an ambiguous-helper qualification. That helper case remains **unreachable and unqualified**.

The genuine native refusal used a separate legitimate private fixture without granting the attempting actor access. Exactly one assessment and one Content attempt returned `permission_denied / note_access_denied`. Public refused/unavailable state and full original owner error/exact input were retained unchanged through cold inspection; carried and cold inspection made zero new assessment/Content calls. There was **no retry or fallback**. Early capture/setup failures remain disclosed. Private fixture values/references are excluded here.

This is an actual **one-read owner refusal**, not pre-dispatch permission detection. The pre-dispatch permission API is absent. Pre-dispatch missing/competing/unsupported non-success has **zero Content reads**; the refusal follows **exactly one attempted authorized owning Content read**. Native owner authority remains intact.

## Judgment, owners and carried scope

Rowan collected the returned support packets; Juniper independently observed actual refusal/count fields and delegating trace. These observations are not Aspen’s pass decision. **Aspen’s qualification judgment is pending at cutoff 4976; main S006 remains underway/unjudged.** Rowan owns the bounded disposition and remaining selected scope; reporting does not add a Root acceptance gate or wait for a future verdict.

S005’s finite guidance-use boundary remains **Qualified and completed** under existing-actor limits. Its complete history retains 34/68 calls, first 22 reader calls transcribed, README 403/local fallback, UI setup assists and beginning/PageDown/end coverage, administrative pre-charter sequence and actual attention activation. Eight cards are explanations, not capabilities.

S004 remains **not completed**, with original unfinished owner scope preserved. All **sixteen full Version 1 outcomes remain incomplete**. Jev remains unselected. Independent GUI/MCP participant acceptance, human/fresh enrollment, long-note behavior, connected harness deployment and full release qualification are not implied by these bounded observations.

## Fixed source boundary and exact history

[Current bounded source metadata](/s006-sources.json) · [Current snapshot](/rearchitecture-status.json) · [Unchanged S006 start snapshot](/development-history/s006-start-progress.json) · [Original S006 plan](/development-history/s006-start-sprint.md) · [Original S006 source metadata](/development-history/s006-start-sources.json) · [Earlier completed/unfinished history](/development/progress-history.html)

Complete native delivery/runtime/participant/negative/refusal/count records are read through cutoff4976. Current integration-owner Plan and named working/maturity references were read later as separately timestamped supplemental observations; this is not atomic event fencing. Exact published source09404cb5/cutoff4900 start JSON/plan/source bytes and earlier S005/S004/S003 histories/agreements are preserved unchanged. Public new metadata contains summary/counts, public source revisions and sequence/title/time, without real private owner/fixture references or private packet data.