Correct canonical skill routes and compose coherent112 source

.agents/skills/check-board/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: check-board
-description: Use when the FS-GG Coordination board looks stale, blockers may have cleared, or planning needs ground truth. Reconcile project fields against live issues and surface judgement calls.
+description: Inspect the explicitly selected FS-GG board against native evidence; explain stale planning, blockers and Unknowns without repairing fields or granting dispatch.
 ---
 
 # check-board (FS-GG)
@@ -48,42 +48,21 @@ fsgg-coord-engine board-v2 inspect --binding-file product-binding.json --report-
 Inspection remains read-only and grants no writes, selection or operation authority. Local-only workspaces
 retain their existing route. This organization preparation changes no product source or acceptance.
 
-### Independently qualified legacy-product reconciliation
-
-The legacy recipes below apply only to a separately qualified legacy-product target and lifecycle.
-Before any product command, verify the actual product target/binding and configured product repository;
-a repository filter alone does not select a board. Refuse missing or organization-default target facts.
-No historical organization live opt-in exists. These recipes cannot repair unavailable V2 evidence or
-select Project 1, including when a caller requests a historical organization scope.
-
-For that verified product target, replace `<configured-product-repo>` with its actual repository:
-
-```bash
-scripts/fsgg-coord reconcile --repo <configured-product-repo> --json
-scripts/fsgg-coord lint --repo <configured-product-repo> --json
-```
-
-`reconcile` is fresh and dry-run unless `--apply` is present. Review its typed mechanical findings and
-the product's existing write authority before the scoped apply; this skill grants no new effect:
-
-```bash
-scripts/fsgg-coord reconcile --repo <configured-product-repo> --apply
-scripts/fsgg-coord reconcile --repo <configured-product-repo> --json
-scripts/fsgg-coord lint --repo <configured-product-repo> --json
-```
-
-A repair is applied only after its fresh post-mutation scan observes the intended fields. Failed or
-stale readback remains failed, never clean. Exit75 retains the reported backoff. Queued writes are not
-landed: queue-wide replay has no repository filter and needs its separately admitted original-effect
-recovery route and genuine target readback. No replay command is authorized by this check-board pass.
-The [offline receipt examples](references/mechanical-reconciliation.md#executable-receipt-examples)
-retain their independently qualified meaning and perform no live writes.
-
-`lint` findings are report-only. Investigate unknown blockers, undeclared paths, unclaimed work and
-outcome/child disagreements without rewriting issues, inventing dependencies or closing work. Report
-product-scoped mechanical results, queued/failed writes, judgement findings and the final fresh result.
-Use [mechanical reconciliation](references/mechanical-reconciliation.md) and
-[judgement findings](references/judgement-findings.md) within that product boundary.
+### Repair and historical product evidence
+
+Inspection is read-only. Before any product command, verify the actual product target/binding and configured product repository;
+Refuse missing or organization-default target facts. No historical organization live opt-in exists.
+A repair needs its own current effect authority, qualified caller and independent readback;
+this entrypoint supplies none. Failed reads retain Unknown rather than selecting an older route.
+
+Historical product reconciliation and the offline receipt examples are retained in
+[mechanical reconciliation](references/mechanical-reconciliation.md) and
+[deep detail](references/deep-detail.md). They are evidence, not a live invocation recipe.
+In particular, queue-wide replay has no repository filter and requires genuine original-effect
+custody and separately admitted recovery. Do not replay a queue from an inspection pass.
+Use [judgement findings](references/judgement-findings.md) to explain discrepancies without
+rewriting issues, inventing dependencies or choosing Status. Current canonical work-board
+owns selection and guarded planning; future controller conversion remains separate.
 
 ## Historical organization examples are data only
 

.agents/skills/check-board/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Check Board"
-  short_description: "Reconcile board state with live GitHub issues"
-  default_prompt: "Use $check-board to reconcile the FS-GG Coordination board against live issue state."
+  short_description: "Inspect board evidence without writes"
+  default_prompt: "Use $check-board to inspect the explicitly selected FS-GG board, explain planning discrepancies and retain Unknowns without repairs or dispatch."

.agents/skills/check-board/references/deep-detail.md

@@ -1,4 +1,9 @@
 
+
+Historical/compatibility evidence only. The current [check-board entrypoint](../SKILL.md)
+is read-only and grants no repair, claim, queue replay or field-write authority.
+Any genuine product repair/recovery requires its own actual qualified caller, target,
+original custody and separately admitted operation. Archived commands are not a live fallback.
 # check-board (FS-GG)
 
 ## Historical and separately scoped mechanics

.agents/skills/check-board/references/mechanical-reconciliation.md

@@ -1,5 +1,10 @@
 # Mechanical reconciliation
 
+Historical/compatibility evidence only. The current [check-board entrypoint](../SKILL.md)
+is read-only and grants no repair, claim, queue replay or field-write authority.
+Any genuine product repair/recovery requires its own actual qualified caller, target,
+original custody and separately admitted operation. Archived commands are not a live fallback.
+
 ## Organization scope is read-only
 
 For prepared .9/.11 organization consumers use only

.agents/skills/continue-from-handoff/SKILL.md

@@ -69,14 +69,15 @@ from worker reports and proposals. Missing temporary evidence remains unknown;
 do not reconstruct results or repeat a consumed operation to fill that gap.
 Apply the selected [forward-recovery boundary](../work-board/references/lifecycle-contract.md#recover-ownership-and-unavailable-history):
 the unavailable historical evidence-search task is closed as unavailable; stop repeated searches and
-requests. Preserve all known original records, 439 inherited reservations, 46 later original records
-and unproved outcomes Unknown, without release or replay. For a genuinely fresh operation, reconcile
+requests. Preserve all known original records and unproved outcomes Unknown, without release or replay.
+Retrieve the actual reservation identities and scope from their existing owning records;
+this generic entrypoint maintains no incident count or cohort ledger. For a genuinely fresh operation, reconcile
 current relevant ownership and remaining capacity, separate generation/writable state/identity/outputs,
 and fence shared destinations against old-actor interference. Missing history is not empty coverage;
 unresolved relevant interference still blocks its operation. Unrelated unavailable history does not
 globally block demonstrated-independent work, and a new ID cannot disguise an uncertain-effect retry.
-Use existing journals/custody, not another recovery ledger. SC2's existing-host container scope does
-not change the separate V3 two-machine portability requirement.
+Use existing journals/custody, not another recovery ledger. Product-specific host or portability
+obligations remain in their actual owning plans rather than this generic recovery entrypoint.
 
 Also recover each pending request's canonical
 channel and original message identity, owner, expected response, response/escalation
@@ -98,16 +99,12 @@ It owns model routing, lanes, telemetry, PR admission, native checks, source int
 separate publication/receiver acceptance boundaries. Do not restore duplicate status or closure upkeep.
 Do not duplicate its lifecycle here.
 
-Invocation authorizes autonomous decisions and completion of the recovered
-Unified Roadmap v3 and the in-scope games outcomes without human intervention, including its necessary source
-changes, PRs, merges, releases, and receiver work within the recorded programme
-scope. Choose implementation details and resolve routine tradeoffs directly;
-do not stop for preference questions or repeated confirmation. This does not
-create missing credentials, bypass protected checks, renew expired grants, or
-replace effect-specific admission. Inspect and obtain valid machine admission
-through the existing route before each concrete effect. Where access or an
-external dependency genuinely prevents an effect, preserve the blocker and
-continue independent ready work; never claim unavailable human feedback.
+Resume only the user's actually authorized recovered scope. Reading or editing this
+entrypoint grants no blanket programme, games, release or receiver authority. Existing
+authorization may permit autonomous completion, but it creates no missing credentials,
+protected approval, runtime capability or renewed original clock. Resolve actual current
+owner, plan and operation admission through work-board before the affected effect.
+Preserve blocked external feedback and continue demonstrated-independent authorized work.
 
 Seek dependency-ready parallel opportunities aggressively within available
 worker, CPU, integrator and hosted-CI capacity. Repair and integrate the existing

.agents/skills/cut-nuget-release/SKILL.md

@@ -1,87 +1,35 @@
 ---
 name: cut-nuget-release
-description: Use when asked to release every stale FS-GG NuGet producer. Audit packed inputs, prepare each coherent set once, publish exact manifest-bound bytes to both feeds, promote, smoke, and reconcile consumer pins.
+description: Explicitly select the maintained FS-GG publishing contract for an owed NuGet release; preserve candidate identity, separate effect admission and selected consumer acceptance.
 ---
 
-# cut-nuget-release (FS-GG)
+# Selected NuGet release entrypoint
 
-Read [publishing-and-deployment](../publishing-and-deployment/SKILL.md) before acting. It owns versioning,
-feed, coherent-set, and registry rules. This skill applies those rules through the repository's current
-pack-once release saga.
+Read [publishing-and-deployment](../publishing-and-deployment/SKILL.md) before acting.
+It owns the current release route, coherent-set versioning, candidate/preflight,
+feed verification, promotion and selected consumer adoption. This compatibility
+name supplies no additional release, retry or recovery authority.
 
-## Invariants
+Resolve the actual producer, changed package inputs, promoted predecessor and genuinely
+owed release. Source delivery, candidate qualification, publication and installed
+adoption remain separate. Do not release every producer merely because this entrypoint
+was invoked; preserve existing publication owners and exact selected operation scope.
 
-- Release only merged, exact-`main` bytes.
-- Never repack in a package publisher. `release-saga-prepare.yml` is the sole pack site.
-- Never reuse a version, move a release tag, overwrite an asset, or bypass OIDC/trusted publishing.
-- The prepared `release-manifest.json` binds every package byte, package id/version, and source SHA.
-- Package publishers download those exact assets, initialize their durable journals, and resume
-  idempotently after a failure.
-- Promote only after every package journal is complete and both feeds expose the complete coherent set
-  with payload identity.
-- Publish first; update canonical registry pins only after both feeds are observed.
+For the hub coherent set, use the selected successor operating window in
+[utel-release-successor](../../../docs/roadmaps/utel-release-successor.md).
+Require its immutable candidate/run/artifact/raw archive and successful `publish=false`
+preflight before a separately admitted publication. Preserve the original bytes,
+journal and independent readbacks; a consumed publication or unresolved send never
+becomes a new operation by changing an identifier or rerunning a workflow.
 
-## 1. Establish what is owed
+Historical release-saga prepare, component-publisher and promote routes remain sealed
+under GS2-08.9. This skill provides no executable legacy recipe or fallback. Missing
+current authority, tooling or feed evidence leaves the affected operation incomplete.
 
-Fetch `main` and tags in every rostered producer without disturbing local changes. Compare each current
-pack input—including staged skills, templates, targets, generated manifests, and runtime sources—with
-the latest successfully published tag. Classify each producer as current, release owed, or blocked.
-Derive dependency order from package references, registry contracts, consumer pins, and dispatch edges.
+Use scoped normal PR delivery for source changes. The global managed `speckit-merge`
+recipe is excluded from FS-GG landing: no branch sweep, direct trunk push/deletion,
+mandatory repack or shared-cache purge. Retain canonical publication and consumer
+owners; do not patch a vendored global skill to repair repository routing.
 
-For the hub coherent set, FS.GG.Coord.Cli, FS.GG.Kit, and FS.GG.Drivers always receive one fresh version.
-Choose the smallest valid SemVer bump under current policy; never infer currency from a project version
-or registry row alone.
-
-## 2. Prepare once
-
-Land version, lockfile, release-note, package-input, and manifest changes through a normal reviewed PR.
-On exact merged `main`, dispatch `.github/workflows/release-saga-prepare.yml` with the fresh version
-and source SHA. Require its Core/CLI/package gates to pass and inspect the draft
-`coherent-set/v<VERSION>` release:
-
-- exactly the expected packages were packed once;
-- `release-manifest.json` binds their sizes and SHA-256 digests to the exact merged source SHA;
-- the draft contains no reused or overwritten assets;
-- all three coherent tags will resolve to that same source SHA.
-
-A preparation run is the reversible rehearsal boundary. Package-publisher workflows have no dry-run or
-local-pack mode.
-
-## 3. Publish and resume
-
-Create/push the coherent tags in the ordering required by the saga, or use the sanctioned recovery
-dispatch with the exact prepared source SHA. Each of `release-coord-engine.yml`,
-`release-kit.yml`, and `release-drivers.yml` must:
-
-1. download the prepared manifest and package bytes;
-2. run `scripts/release-saga-ci.sh init` for its package and exact source SHA;
-3. publish the prepared package to GitHub Packages;
-4. record the GitHub observation in its immutable package journal;
-5. authenticate to nuget.org through OIDC trusted publishing;
-6. publish the same prepared payload and record the public observation.
-
-On failure, repair the cause and rerun the same publisher against the same manifest. The journal must
-make completed steps idempotent and reject identity, source, version, or byte drift.
-
-## 4. Promote and verify
-
-Dispatch `.github/workflows/release-saga-promote.yml` only after all three journals exist. Require it to
-verify the manifest, source SHA, coherent tags, complete journals, both feeds, and monotonic channel
-ordering before promotion. Then:
-
-- download each package from both feeds and compare payload entries, excluding nuget.org's
-  `.signature.p7s`;
-- install the public coord engine in a clean tool home and run command-contract plus safe live reads;
-- materialize Kit/Drivers in a clean consumer and verify expected files, modes, digests, and versions;
-- update consumer pins and `registry/dependencies.yml` through the sanctioned generated projections;
-- merge the registry/pin PR, fetch exact `main`, and rerun feed, source, pin, and coherent-set checks.
-
-## Definition of done
-
-Done means the complete expected set is available from both feeds, payload-identical, promoted, adopted,
-and pinned on merged `main`; all package/tag/source identities agree; fresh-consumer smoke passes; and
-no stale bot PR/branch remains. A tag, one green publisher, one feed, or an updated registry alone is
-not completion.
-
-Report exact version, source SHA, PRs, tags, workflow runs, journals, feed observations, promotion,
-consumer smoke, and merged registry/pin evidence. If any part remains, report the release as incomplete.
+Report the actual selected source/candidate, operation result and remaining acceptance.
+A package, tag, registry digest or green source PR alone does not complete adoption.

.agents/skills/cut-nuget-release/agents/openai.yaml

@@ -1,7 +1,7 @@
 interface:
   display_name: "Cut NuGet Release"
-  short_description: "Release every stale FS-GG NuGet producer"
-  default_prompt: "Use $cut-nuget-release to audit and publish every stale FS-GG NuGet producer in dependency order."
+  short_description: "Select the maintained release contract"
+  default_prompt: "Use $cut-nuget-release to resolve the owed producer release through canonical publishing-and-deployment, preserving original bytes and separate effect admission."
 
 policy:
   allow_implicit_invocation: false

.agents/skills/padd-item/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: padd-item
-description: Add or reuse a request for the explicitly configured product through canonical intake; preserve qualified legacy-product compatibility.
+description: Add or reuse a request for the explicitly configured product through canonical native intake; preserve exact scope and keep retired legacy recipes non-executable.
 ---
 
 # Configured product intake
@@ -15,6 +15,7 @@ Reuse an existing semantic match. If another repository owns the request, use
 
 For selected V2 product scope, do not run legacy intake/reconcile/apply/add/flush. Return created/reused
 identity and exact pending effects, or a concrete draft when the native operation is unavailable.
-Only an independently qualified legacy PRODUCT invocation may load
-[legacy-product intake](references/legacy-product-intake.md); it cannot authorize organization operations.
+The [retired legacy-product appendix](references/legacy-product-intake.md) records its disposition,
+not an executable alternative. No legacy receiver is selected here; any future exception needs its
+actual supported target, installed caller and separately admitted effect authority.
 File the requested work; do not implement it. A clearing invocation returns to work-board after intake.