docs: record four-stage orchestration adoption and native evidence by seathatflowsinourveins · Pull Request #904 · native-agent-stack/native-agent-stack

seathatflowsinourveins · 2026-10-09T00:07:20Z

Scope

Records the current host-wide activation of Claude Teams and workflows, leaves role-scoped loading owner-gated with its inverse, pins the payload selected by the supported AO install recipe, and retains the exact frozen snapshot used in invocation comparisons. The four-stage adoption record keeps all four outcomes DEFER pending quality, memory and fresh owning-role evidence.

Base commit: aba02ec3456d383bcc2fc72883db098f9be7918a.
Lane: lane:foundation.
Owned paths: the three docs/decisions/2026-10-08-orchestration-* records and docs/decisions/extra/adoption-cc-2ebc30ce3527d025.json.

SOTA sources

aannoo/hcom v0.7.28, b2a7c192003e7fd67ed93265289e4ac36276f965: README, native event projection, common/Codex hooks and kill semantics. The canonical source list is aligned across the records; the three hashed measurement files retain their distinct scope.
anthropics/claude-code v2.1.295, 602df92bf481ed904533e95c09f740f40aab5aed: CHANGELOG.md, section 2.1.178 implicit teams. Installed build 07e8f67ea328, executable SHA256 4503bfe11a6c7fcc1e0b39b5e0d347c04248f750b03b0977b3ad6b531fe6f358, independently pinned selection/coordination/session spans in the native receipt. CC-verified user-scope activation is recorded separately from the proposed role change and natural selection.
OrchestratorInc/agent-orchestrator v0.13.5, c95ae361eee48d33c2f443c6d2fe69c445f548a8: scripts/setup-self-hosted.sh:90-107,141-170 supported local bundle and native extraction. Installer SHA256 463366604181029d8cdb16cbc1a5d8771eaf4bc498594b09759a5b42b79bee7d; exact Linux x64 AppImage asset 621487952, SHA256 5f4262c1b5fad5e63c0b1291d5a2e4c878c953496f5ca430edfe4c061987bf28. The recipe enforces these hashes before deriving a separately checksummed resources archive for --bundle; execution remains owner-gated.
Both designated delta artifacts at 0ad947f6fd5023b2a21eadae2837349d4b313886 are bound in review_sources: CC 69119fd821ddab7fda786a8b374ccfa89db43aa43cd81edb49b9fd5714860357; GPT bb3bfe43911116f1dfb8721bc36c6cedc32c90d6f235287f511acbbc4ad14735.

Evidence-class table

Claim	Evidence class	Command / receipt
Host-wide Teams/workflows activation	`native_proven` by the CC	CC delta artifact; settings locators and CHANGELOG contract in `native_evidence.claude_host_activation`. Role scoping and its inverse remain proposed.
Fixed AO payload selection	`source_review`	Pinned installer plus public v0.13.5 asset metadata; both prospective recipes pass `bash -n`. The AppImage digest pins the upstream payload; the derived tar has a separate owner-retained digest.
Frozen CC comparison source retained	`local_integration`	Exact 60,603-byte file, SHA256 `2ebc30ce3527d025dad39d65293d3020620192c4f0d2987cf3988461254836a0`, committed under `docs/decisions/extra/` and copied to the new-head packet's `extra/`.
Existing counters, denominators, source spans and PSS observations	Scoped retained evidence	Compared against the previous head and kept unchanged. Invoke counts remain baseline only; component costs and organic acceptance remain open.

The record also carries the six optional clarifications: small-PR bounds, CC deployment stop guard and owned-service inverse, consistent hcom source scope, refresh procedure pending #878, portable state root, and the stored-delivery-field predicate including empty lists. No stronger acceptance class is inferred from these corrections.

Local commands run

nice -n 10 ionice -c2 -n7 timeout 600 env PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. python3 -B -m unittest tests.test_adoption_contract
rc=0; 13 passed / 13 run
nice -n 10 ionice -c2 -n7 timeout 600 env PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. python3 -B -m unittest tests.test_adoption_bootstrap
rc=0; 76 passed / 80 run; 4 skipped
nice -n 10 ionice -c2 -n7 timeout 600 env PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. python3 -B -m unittest tests.test_validate
rc=0; 68 passed / 68 run
nice -n 10 ionice -c2 -n7 timeout 600 python3 scripts/validate.py
rc=0
git diff --check
rc=0

The two generic documentation reader modules and validator module are publication/regression checks. They do not establish organic adoption or execute the AO install recipe. The initial publication check rejected an absolute personal state path; the final record uses the portable ~/.local/state/native-agent-stack root and passes validation.

Decision record

Four-stage record, decision JSON, native evidence, and retained CC snapshot. Quality/G5 evidence selects one current choice per role slot; invocation volume does not select winners. AO extends the v0.13.4 history and its bounds.

Both new-head micro reads, complete CI, the co-op's designated pre-cue tool output and the CC's cue remain required. G5 bindings, component PSS, owner role scoping/install and fresh natural-task acceptance remain open. Landing belongs to 5f on the CC's cue.

Host evidence

The retained evidence is under docs/decisions/; this revision adds no evidence/hosts/ receipt.

Checklist

Record-only scope; native observations and frozen comparison counts retained.
Supported upstream payload route and full source/content pins recorded.
Publication validation and appropriate targeted modules completed.
Both micro reads, complete CI, designated pre-cue output and CC cue.
G5, component memory, owner decisions and fresh owning-role acceptance.

chatgpt-codex-connector · 2026-10-09T04:35:37Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-09T04:40:59.378263Z	`836195c`	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 836195c432

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-10-09T04:40:56Z

+    "fixture_ref": "evidence/receipts/ao-bounded-effect-20261008.json",
+    "fixture_scope": "Synthetic input with native AO engine, not organic/provider acceptance"


Point the retained execution at its matching receipt

When a consumer follows fixture_ref to verify this retained execution, the referenced receipt describes a different experiment: evidence/receipts/ao-bounded-effect-20261008.json records install exit 0, native stop exit 0, no started runtime, and zero model calls (lines 23, 180-191), whereas this object claims rc 1, three quota-failed proposal turns, and native stop 137. Its file SHA256 is also fc3e64..., not the receipt_sha256 above. Separating the synthetic fixture reference from the matching retained execution receipt, or linking the actual receipt, is necessary to keep these outcomes reproducible and correctly attributed.

AGENTS.md reference: AGENTS.md:L5-L7

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-09T04:40:57Z

+            "command": "curl -fsSL https://github.com/aannoo/hcom/releases/download/v0.7.28/hcom-installer.sh | sh",
+            "source": "Vendor README installer method, fixed release asset at v0.7.28",
+            "installer_sha256": "75c1560785799881c265ae02ccb247bc6f0742835d716452f7d82071fe8bd638",


Verify the hcom installer before executing it

When an owner follows this documented clean-install command, the recorded installer_sha256 is never checked: the release script is streamed directly into sh, so the installation executes whatever bytes the URL returns rather than the pinned artifact represented by the adjacent digest. Fetch the installer to an owned file, verify 75c156..., and only then invoke it, as the repository's existing verified-install flow does; otherwise this record's claimed pin provides no integrity or reproducibility guarantee.

AGENTS.md reference: AGENTS.md:L5-L7

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-09T04:40:57Z

+    "selection_sql": [
+      "SELECT id,timestamp,instance,life_action,life_by,life_batch_id,life_reason='killed',json_extract(data,'$.tag'),json_extract(data,'$.tool'),json_extract(data,'$.instances'),json_extract(data,'$.snapshot.tag'),json_extract(data,'$.snapshot.tool'),json_extract(data,'$.snapshot.created_at'),json_extract(data,'$.snapshot.session_id'),json_extract(data,'$.snapshot.name'),json_extract(data,'$.launched'),json_extract(data,'$.failed') FROM events_v WHERE type='life' AND id<=? ORDER BY id;",
+      "SELECT name,tag,tool,created_at,session_id FROM instances ORDER BY name;",
+      "SELECT id,timestamp,instance,status_val,json_extract(data,'$.session') FROM events_v WHERE type='status' AND timestamp>=? AND timestamp<? AND id<=? ORDER BY id;",
+      "SELECT id,timestamp,instance,msg_from,msg_sender_kind,msg_intent,msg_reply_to,msg_reply_to_local,msg_delivered_to FROM events_v WHERE type='message' AND timestamp>=? AND timestamp<? AND id<=? ORDER BY id;"


Retain the lifecycle window filter used for these totals

When another reader replays the recorded selection_sql, the lifecycle query returns every life event up to ID 65430, while the receipt claims lifecycle totals for only [2026-10-08T00:31:25Z, 2026-10-09T00:31:25Z) and records no post-query filtering step. The status and message queries include the window predicates, but this one does not, so the listed SQL and parameters cannot reproduce the 410 lifecycle events or the ready/stopped aggregates. Retain the actual post-query aggregation rule, or use a separate window-bounded lifecycle query for those totals while preserving the full-history query needed for lifespan attribution.

AGENTS.md reference: AGENTS.md:L6-L7

Useful? React with 👍 / 👎.

docs: record host orchestration adoption decisions

021e1a8

seathatflowsinourveins added the lane:foundation Foundation lane: Claude/Codex setup, hosts, memory, RAG, research, workers label Oct 9, 2026

docs: bind orchestration stages to native sources and role evidence

0ad947f

seathatflowsinourveins changed the title ~~docs: record orchestration KEEP/WIRE decisions and native evidence~~ docs: record four-stage orchestration adoption and native evidence Oct 9, 2026

docs: correct orchestration scope and pin retained evidence

836195c

seathatflowsinourveins marked this pull request as ready for review October 9, 2026 04:35

seathatflowsinourveins merged commit d641790 into main Oct 9, 2026
36 checks passed

seathatflowsinourveins deleted the codex/ns2604-orchestration-adoption-record-20261008 branch October 9, 2026 04:35

chatgpt-codex-connector Bot reviewed Oct 9, 2026

View reviewed changes

Navigation Menu

Repository navigation

docs: record four-stage orchestration adoption and native evidence - #904

docs: record four-stage orchestration adoption and native evidence#904
seathatflowsinourveins merged 3 commits into
mainfrom
codex/ns2604-orchestration-adoption-record-20261008

seathatflowsinourveins commented Oct 9, 2026 •

edited

Loading

Uh oh!

Uh oh!

chatgpt-codex-connector Bot commented Oct 9, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

		"fixture_ref": "evidence/receipts/ao-bounded-effect-20261008.json",
		"fixture_scope": "Synthetic input with native AO engine, not organic/provider acceptance"

Conversation

seathatflowsinourveins commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Scope

SOTA sources

Evidence-class table

Local commands run

Decision record

Host evidence

Checklist

Uh oh!

Uh oh!

chatgpt-codex-connector Bot commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

seathatflowsinourveins commented Oct 9, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 9, 2026 •

edited

Loading