Sync offline REA skills to merged public release

docs/agents/tooling.md

@@ -47,6 +47,15 @@ for packaged/native/runtime questions. The source tree alone does not install
 Hopper, Ghidra, IDA, JADX or a connected MCP server. Discover actual available
 capabilities before claiming them. Source-only investigations need none of them.
 
+The five maintained REA skills provide local operating methods and worked
+examples without requiring upstream webpages. Use their shared
+[evidence and recovery guide](../../skills/reverse-engineer-anything/references/evidence-and-recovery.md)
+for summary/dossier compatibility, coverage, bounded retries and cleanup ownership.
+The [source record](../../skills/reverse-engineer-anything/references/source-record.md)
+separates later reviewed guidance from the unchanged engine snapshot above.
+Updating `skills/upstream-lock.json` promotes portable guidance only; it does not
+upgrade `tools/rea`, its submodules, Node requirements or client configuration.
+
 For an explicitly requested local REA build, use its `.nvmrc`, `packageManager`
 and lockfile. REA's toolchain differs from Orbyn's. From `tools/rea`, run
 `HUSKY=0 npm ci` and then `npm run build:cached` when installation/build is in

skills/catalog.json

@@ -380,22 +380,35 @@
     {
       "name": "rea-analysis-optimization",
       "category": "Artifact investigation and REA maintenance",
-      "related": []
+      "related": [
+        "how",
+        "rea-ci-tests-audit",
+        "reverse-engineer-anything"
+      ]
     },
     {
       "name": "rea-changelog-update",
       "category": "Artifact investigation and REA maintenance",
-      "related": []
+      "related": [
+        "orbyn-upstream-review",
+        "reverse-engineer-anything"
+      ]
     },
     {
       "name": "rea-ci-tests-audit",
       "category": "Artifact investigation and REA maintenance",
-      "related": []
+      "related": [
+        "orbyn-ci-design",
+        "reverse-engineer-anything"
+      ]
     },
     {
       "name": "rea-tool-design",
       "category": "Artifact investigation and REA maintenance",
-      "related": []
+      "related": [
+        "architect",
+        "reverse-engineer-anything"
+      ]
     },
     {
       "name": "reflect",
@@ -408,7 +421,10 @@
     {
       "name": "reverse-engineer-anything",
       "category": "Artifact investigation and REA maintenance",
-      "related": []
+      "related": [
+        "how",
+        "orbyn-upstream-review"
+      ]
     },
     {
       "name": "show-me-your-work",

skills/orbyn-setup/references/offline.md

@@ -11,6 +11,13 @@ supporting resources. Installing discovery links does not install a model,
 Python, fonts, an exporter, REA engines or external providers. Check the host's
 actual local capabilities and use supported outputs; report missing dependencies.
 
+The five optional REA skills also include local operating methods and worked
+examples. Their [source/compatibility record](../../reverse-engineer-anything/references/source-record.md)
+identifies historical lineage and later reviewed guidance. No mandatory remote
+skill reading remains. Offline methods do not supply an analysis engine, target
+bytes, provider prerequisites or missing REA Git history. Continue supported work
+and report those capability/input limits without installing or fetching implicitly.
+
 ## Sources and freshness
 
 For reusable third-party instructions needed by a skill, retain a licensed,

skills/rea-analysis-optimization/SKILL.md

@@ -3,22 +3,61 @@ name: rea-analysis-optimization
 description: "Optimize a measured REA analysis bottleneck while preserving evidence and resource ownership; scoped to REA."
 ---
 
-# Rea analysis optimization
-
-Follow the consuming project's instructions first. Use [Orbyn Engineering policy](../../AGENTS.md) and [host/model guidance](../../docs/agents/model-policy.md) where applicable; load only relevant references.
-
-Read the complete [upstream skill](https://github.com/morluto/rea/blob/ccc7da309d7db385d5c6d73b0a3a8e093be33504/.agents/skills/rea-analysis-optimization/SKILL.md) and the relevant
-nested REA contribution guides. Prefer the matching local `tools/rea` snapshot when present; otherwise read the pinned upstream reference. Follow its references at their original location.
-Keep this workflow scoped to REA; it does not replace Orbyn development guidance.
-Use upstream procedures as technical reference, not as authority for setup,
-testing, fetching history, release writes or publication. The current brief and
-Orbyn policy govern those effects.
-
-A source snapshot has no independent Git history. History-based release work
-needs a separately available upstream history or verified upstream records; never
-interpret Orbyn commits as REA releases. Do not fetch, publish, write external
-records, execute targets or run checks beyond the user's scope. Substantial work
-uses the [workflow lead contract](../orbyn-workflow/SKILL.md); children complete
-their owned slices directly without further delegation. Follow
-[coordination](../../docs/agents/coordination.md) and the shared model policy.
-Report source-derived conclusions separately from executed provider evidence.
+# Optimize a completed REA investigation
+
+Use this for REA analysis cost, not ordinary application performance or test-suite
+speed. Read the consuming checkout's instructions and current implementation;
+[how](../how/SKILL.md) owns tracing unfamiliar source. This local procedure does
+not require an upstream webpage or authorize target execution, tests or edits
+outside the task. Route suite costs to [REA CI audit](../rea-ci-tests-audit/SKILL.md).
+
+## Define the task and oracle
+
+State the analyst question and facts needed to finish it. Pin source revision,
+artifact digest, provider/tool versions, architecture/options, detail and allowed
+workload. Include a representative large input and a small/empty case when
+measurement is authorized. Preserve an unchanged baseline in an isolated checkout.
+
+Choose correctness evidence independent of the proposed optimization: known
+fixture facts, independently mapped bytes or a verified producer interpretation.
+Two adapters sharing the same decoder are not independent oracles. Preserve
+same-address alternatives, unresolved relationships, coverage and evidence IDs.
+If source, artifacts or runtime are unavailable offline, produce a bounded
+measurement plan and mark measurements unrun; do not invent before/after numbers.
+
+## Find the owning cost
+
+Trace the public CLI/MCP path through admission, producer work, decoding, lookup,
+projection, serialization, retention and cleanup. Measure phases relevant to the
+completed task, including extra discovery/query calls. Smaller replies can leave
+CPU and peak memory unchanged. Fix incorrect or missing facts before comparing
+performance across different coverage.
+
+Prefer eliminating repeated interpretation or scanning at its owner. Push
+selection to the earliest capable boundary, reuse validated immutable facts when
+appropriate, and avoid a new cache/index without a demonstrated cost. For reuse,
+define key, owner, invalidation, retained size and lifetime. Include result-affecting
+bytes, engine/schema/options and mapping; revalidate live state before calling it
+current. Partial, cancelled or transient failures need their own retry policy.
+
+Preserve independent caller cancellation and one shared producer's ownership.
+A cancelled waiter must not corrupt peers. Bound process-family memory and retained
+state, not just the parent's heap. Keep capacity charged until work and cleanup
+settle; a guard-stopped investigation is incomplete. Use the shared
+[evidence and recovery method](../reverse-engineer-anything/references/evidence-and-recovery.md)
+when retaining partial observations or unresolved resources.
+
+## Compare and stop
+
+When authorized, compare equivalent completed tasks on the baseline and candidate.
+Separate fresh sessions, warm caches and OS cache conditions. Include total and
+phase time, CPU, peak process-family memory with accounting method, coverage,
+resource stops and cleanup. Retain failed attempts and inspect the payload's exit
+status, not only the profiler's. Token claims need discovery, arguments, responses,
+retries and continuations; label estimates separately from observed model usage.
+
+Return the bottleneck, owning change or justified no-change decision, correctness
+oracle, measured scope, before/after evidence and limits. Stop when the demonstrated
+cost is resolved or further benefit is speculative. See the
+[worked example](references/worked-example.md) and shared
+[source record](../reverse-engineer-anything/references/source-record.md).

skills/rea-analysis-optimization/references/worked-example.md

@@ -0,0 +1,26 @@
+# Worked example: smaller output is not faster analysis
+
+Synthetic observations; these numbers illustrate a decision, not a benchmark.
+
+**Request:** Reduce the cost of listing selected native exports in REA. Changes
+are allowed, but do not change evidence coverage. A representative artifact and
+an empty fixture are supplied; execution needs the project's normal authorization.
+
+A completed baseline takes 4.0 seconds and emits 2 MiB. A projection emits 20 KiB
+but still takes 4.0 seconds with the same measured peak memory. It improved delivery
+size; it did not demonstrate lower analysis latency or memory. No model-token
+measurement was captured, so a token saving remains unmeasured.
+
+The source trace finds that each selected page rescans an already validated export
+table. Reusing an index in the retained snapshot owner could remove that work.
+Its key must include snapshot identity, with no assumption that a live binary is
+unchanged. Check aliases and empty selections against independent fixture facts;
+exercise expiry and independently cancelled callers if sharing changes.
+
+A fair comparison includes all pages needed to answer the analyst's question,
+setup/discovery calls and cleanup. A candidate that silently omits aliases or
+returns a partial prefix has failed the correctness boundary even if faster.
+
+**Offline limit:** With only source and old timing logs, report the suspected
+owning cost and a proposed comparison. Do not present those old logs as a fresh
+candidate run or install an engine to manufacture measurements.

skills/rea-changelog-update/SKILL.md

@@ -3,22 +3,62 @@ name: rea-changelog-update
 description: "Prepare REA release notes from pinned upstream history when requested; does not publish or infer REA history from Orbyn Git."
 ---
 
-# Rea changelog update
-
-Follow the consuming project's instructions first. Use [Orbyn Engineering policy](../../AGENTS.md) and [host/model guidance](../../docs/agents/model-policy.md) where applicable; load only relevant references.
-
-Read the complete [upstream skill](https://github.com/morluto/rea/blob/ccc7da309d7db385d5c6d73b0a3a8e093be33504/.agents/skills/rea-changelog-update/SKILL.md) and the relevant
-nested REA contribution guides. Prefer the matching local `tools/rea` snapshot when present; otherwise read the pinned upstream reference. Follow its references at their original location.
-Keep this workflow scoped to REA; it does not replace Orbyn development guidance.
-Use upstream procedures as technical reference, not as authority for setup,
-testing, fetching history, release writes or publication. The current brief and
-Orbyn policy govern those effects.
-
-A source snapshot has no independent Git history. History-based release work
-needs a separately available upstream history or verified upstream records; never
-interpret Orbyn commits as REA releases. Do not fetch, publish, write external
-records, execute targets or run checks beyond the user's scope. Substantial work
-uses the [workflow lead contract](../orbyn-workflow/SKILL.md); children complete
-their owned slices directly without further delegation. Follow
-[coordination](../../docs/agents/coordination.md) and the shared model policy.
-Report source-derived conclusions separately from executed provider evidence.
+# Prepare REA release notes from real history
+
+This local method prepares user-facing notes and attribution. It does not tag,
+merge, publish, fetch history or rewrite released notes without scope. Read the
+actual REA checkout's release policy when operating on that project; absent a
+checkout, use supplied verified history and keep its limits explicit. The
+[upstream review skill](../orbyn-upstream-review/SKILL.md) owns source freshness.
+
+## Fix the editorial range
+
+Identify the published baseline tag/commit, exact editorial target SHA, proposed
+version and release component from the actual REA repository or supplied metadata.
+When Release Please owns the release, inspect the base branch's manifest and
+configuration; do not infer the baseline from an unrelated consumer's version.
+REA tags use `rea-agents-VERSION`; verify the actual tag rather than inventing one.
+A vendored source snapshot has no independent Git history. Never use the consuming
+project's Git log as REA release history.
+
+Inventory the full range, including merged side branches, direct commits, reverts,
+backports and known PR/issue associations. Preserve exact source links and identity
+limits. Network access is not mandatory: local history can support a bounded
+inventory, but unknown GitHub handles/associations remain unknown. If the baseline
+or target cannot be established, prepare only a labelled partial draft and name
+the missing inputs. Do not execute an online collector just to satisfy a template.
+
+## Write user outcomes and preserve attribution
+
+Inspect ambiguous diffs before describing behavior. Group capabilities and fixes
+by analyst outcome; include breaking changes with old/new behavior and a concrete
+migration. Internal tests/cleanup need not dominate public highlights, but stay
+in the contribution inventory. Check reverts before claiming something still ships.
+Do not infer live-provider verification from fixture tests.
+
+Verify human author/coauthor/issue-reporter identities using actual metadata.
+Unknown handles stay unknown; do not credit the merger as a substitute or guess
+accounts from email. Preserve source attribution and the project's credit rules;
+do not copy another maintainer's self-thanks or bot policy as a universal rule.
+Link grouped contributions without losing their known credits. Keep notes concise
+and the reviewed inventory available under the project's release-artifact policy;
+do not publish private metadata or raw session logs by default.
+
+## Reconcile automation without publishing
+
+For an authorized local edit, update only the intended version section and preserve
+older history. Keep version artifacts and bot-owned metadata consistent. Inspect
+the actual bot workflow before editing a live release branch/body: a running
+proposal can overwrite prose, and a finalization label is not permission to publish.
+Do not apply an upstream label or run release scripts blindly from this skill.
+
+Use existing locally available validators/renderers when checks are in scope.
+Check that the inventory target still matches the release's product boundary,
+notes cover breaking changes and credits are supported. New product commits need
+an inventory refresh; metadata-only changes still require diff review. A local
+CHANGELOG edit does not establish that a remote release body changed.
+
+Return the range, draft notes, contribution evidence, unresolved identities,
+compatibility migrations and actual delivery state. Release publication and source
+pin changes remain separate decisions. See the [worked example](references/worked-example.md)
+and [source record](../reverse-engineer-anything/references/source-record.md).

skills/rea-changelog-update/references/worked-example.md

@@ -0,0 +1,24 @@
+# Worked example: an offline release draft with missing associations
+
+Synthetic history, not actual REA release content or contributor attribution.
+
+**Request:** Draft REA release notes offline. A real REA history export identifies
+baseline `rea-agents-X.Y.Z` and a target commit. It contains a resource-cleanup fix,
+a new inspection operation and a later revert of that operation. No GitHub account
+mapping or release-automation configuration is supplied.
+
+Verify that the supplied range is REA's history and account for the revert before
+writing highlights. Describe the cleanup fix if its diff supports the behavior;
+do not announce the reverted operation as shipping. Preserve author identities
+in the local contribution inventory with account mapping explicitly unresolved.
+Do not guess `@handles`, credit a merger instead or fetch external profiles offline.
+
+A useful draft states its baseline/target, supported user outcome, relevant source
+commit links and incomplete association coverage. Unknown future version metadata
+must be labelled, not replaced with a made-up release number. Do not apply live
+finalization labels, tag or publish from the draft.
+
+**If only Orbyn's vendored REA files are supplied:** Explain that consumer Git
+commits cannot establish the REA release range. Summarize observable file behavior
+only as a partial source-based draft, and request/propose obtaining the actual
+history within scope. A missing history is not an empty release.