Skip to content

Prove public-target subscription support and native App eligibility #173

Closed as not planned

Description

@HemSoft

Description

Planning backlog under the alpha specification. The owner's October 19 merge-mission invocation authorizes scoped offline implementation and guarded merges. The retired project remains frozen; credentials, App installation and live activation retain separate authority and evidence gates.

The alpha specification requires both Codex and Claude subscription adapters and a new custom App publishing identity. Unattended Codex subscription use against this public repository remains unresolved. Native-picker requests failed in the model-free experiment and have been replaced by Franz's explicit comment plus automatic polling-trigger decision. Produce an evidence-backed gate report before live activation.

Why

Running on a private host does not establish a public-target auth exception. Generic Bot fields also do not prove the requested App can be selected, discovered and re-requested.

Goal

Each eligibility gate has an explicit pass, fail or blocked result with primary/account-specific evidence. Offline development can continue independently once authorized; live activation remains blocked until applicable gates and owner authority are satisfied.

Trigger decision — October 21, 2026

Franz chose @sfl review plus automatic PR-open/head-push review on October 21, 2026 at 01:19 EDT, via Discord relayed by the Chief. Implement both offline with fake transports and normal review/merge gates. Discovery stays polling-based; no webhook, machine user, live inference, subscription use, deployment or paid usage is authorized by this decision. Native-picker requests are no longer required. PR #222 records the failed native experiment; #221 is closed unmerged.

Acceptance Criteria

  • Establish exact support for owner-only, unattended public-PR review with both pinned, unmodified official CLIs and subscription auth, including refresh/expiry. Address OpenAI's public-target exclusion in its CI auth guide with applicable documentation or written confirmation, not a successful request or private-host assumption. Record Claude's authentication and credential-use requirements.

  • Document subscription entitlements, model access, actual auth-mode checks and enforceable prevention of paid overflow for each account. Never include tokens or auth files in evidence.

  • Run the authorized model-free App experiment with the selected installation covering exactly ID 1169772257. New App/Bot identity, permissions and COMMENT authorship passed; UI and API native request/re-request creation failed; dependent native lifecycle gates could not be exercised. Model-free App reviewer identity experiment — close unmerged #221 closed unmerged. Preserve the dated evidence in Record new App native-reviewer eligibility results #222. Native eligibility is superseded as a requirement by the explicit replacement-trigger decision; validate new trigger attribution/downtime/A-B behavior offline under Build the durable alpha controller and scoped review publisher #176 and model-free/live only with applicable authority.

  • Record required permissions, new App/installation identity and cleanup steps. Never reuse the retired hs-buddy identity. No model inference during the identity experiment.

  • Preserve blocked outcomes without changing visibility, identity, provider, billing mode or repo scope. The only authorized trigger change is @sfl review plus automatic PR-open/head-push polling. A support-question evidence package may be prepared; sending it requires owner authorization.

  • Run the model-free experiment on public throwaway Model-free App reviewer identity experiment — close unmerged #221 and close it unmerged. The Chief supplied new App5406096/installation173628126 and key under the separate October20 20:28 EDT owner authorization. No inference or alpha activation occurred; credential recreation/broader installation remain separately gated.

Definition of Done

  • Commit the dated gate report, reproducible experiment steps and sanitized evidence under the new alpha documentation. Update the source issue with outcomes and exact remaining needs.
  • Identify which gates are documentary, account-specific, model-free and credential-bearing. No live inference until support, isolation and activation gates pass.

Validation

  • Independently follow each documented model-free step as HemSoft after authorization; inspect resulting GitHub request/event/review IDs and re-request behavior.
  • Verify cited provider guidance covers the exact unattended public-target case and the official-client auth lifecycle. Mark ambiguity blocked rather than pass.
  • Check Actions remains disabled and no legacy service, credential or org-wide installation was restored. Embed rendered picker/request/completion screenshots as uploaded GitHub attachments in the implementation PR, with a recording where practical; state why if omitted.
  • Record validation output/evidence at the exact PR head SHA. For code changes, the reviewer reruns applicable commands locally; GitHub Actions CI is currently disabled.

Reporting reconciliation — October 21, 2026

The checked reporting items are proved by the October 21 model-free report and its linked sanitized API/UI/closure receipts, merged in PR #222. They document new App 5406096, Bot 345646943, selected installation 173628126, exact repository 1169772257, read contents/metadata plus write pull_requests, model-free COMMENT authorship and #221 closed unmerged. App/key retention follows the owner's separate decision; no uninstallation or credential revocation is claimed.

Applicable unattended public-target permission, account-specific worker entitlement and enforceable paid-overflow evidence remain unproved. This checklist reconciliation grants no provider use or activation and leaves #173 open. The native-picker result remains historical; the selected comment/automatic polling triggers retain their independent gates.

Merged activation checklist, October 21, 2026

PR #233 merged the dated six-gate checklist as ab260ad2d6883ddc4a22446a58ebfb93161296c6, after clean required review of e4c4a82a3b2178a4e9346c2955d0cdcbd3f00c38 and five addressed threads. It separates implemented mechanisms from unproved physical and live evidence and prepares the selected-trigger provider questions without sending them.

Evidence classes are explicit: provider applicability needs documentary guidance or a scoped written answer; entitlement/auth/renewal/paid overflow need account-specific evidence; synthetic containment and canned App/UI probes are model-free; actual subscription evaluation and live service use are credential-bearing and separately gated. The completed App experiment supplies only its dated identity/COMMENT/picker evidence. No blanket account or activation grant follows.

The six live gate groups remain open: applicable provider permission; account/auth/paid-overflow evidence; complete aggregate containment/recovery; actual comparative evaluation plus independently qualified trusted release/plan and pilot; final selected App/trigger/rendering qualification; separate explicit owner activation. PR #234 merged as 917d854 after a clean exact-head review. Its four controlled synthetic supervisor/driver-loss cases prove original-unit stop/recovery at those points, not arbitrary host crashes or aggregate qualification. The original host noswap/sealed-exec failures remain failures. PR #235 merged as 00d6a708 after a clean exact-head review and all 15 threads resolved. It delivers an inert measured VM prerequisite diagnostic and a design for a distinct VM source adapter; it grants no activation. This issue remains open.

PR #236 merged as 0f333755 after clean exact-head review, all four threads resolved and verified delivery/cleanup. PR #237 merged as 60d64113 with the guest-role original-descriptor fix, clean exact-head review, full vet/race and six passing synthetic-client VM runs covering normal completion and cancellation. PR #238 merged as 39bc435 after a clean exact-head review, zero threads and verified delivery/cleanup. It binds two inert diagnostics to their captured original unit. Its 48 Python tests, actual escaped-child/replacement checks and KVM fixture pass; the inert host sealed-exec comparison still reports incompatibility. Production source-VM acquisition and startup integration, aggregate qualification and separate live gates remain open. Source work is not exhausted.

Activity

  1. HemSoft commented on Oct 19, 2026

    @HemSoft
    CollaboratorAuthor

    The evidence-report stage merged in PR #179, with clean current-head Codex review. Full live eligibility remains blocked. Checked two routes for Codex: cached-account CI excludes public targets; general script/workspace-token documentation supplies no applicable account/public-target proof. Checked two routes for App eligibility: native Bot schema does not prove picker enrollment; authenticated named-App lookup is 404 and org installation inventory has no retired SFL entry. CLI status proves existing interactive subscription auth only, not worker entitlement or disabled paid overflow. Exact remaining needs and the model-free experiment are in the merged report. No credentials, App, billing settings or runtime were created or changed. Offline implementation proceeds under the current merge-mission instruction; this issue stays open for its full acceptance contract.

  2. HemSoft commented on Oct 19, 2026

    @HemSoft
    CollaboratorAuthor

    Eligibility remains blocked after the fresh read-only revisit. The target is public, repository ID 1169772257, Actions disabled, zero workflows. No alpha App installation exists in the authenticated organization inventory. No App was installed and no worker credentials or billing settings were changed.

    Two support routes were checked: cached-account CI guidance expressly excludes public/open-source targets; general scripted mode plus workspace-token documentation does not establish this owner's entitlement or an exception for the exact unattended public-target service. This is missing applicable support evidence, not a claim that every form of personal scripted use is prohibited.

    Two App routes were checked: native bot request schema does not prove arbitrary custom-App picker eligibility; authenticated App lookup/organization installation inspection found no usable alpha identity. Existing connected Codex and Vercel installations are not substitutes for a new scoped reviewer App.

    Sanitized CLI status still reports Codex ChatGPT auth and Claude first-party Pro auth. That route and documented subscription token support establish mechanisms, not separate worker credentials or disabled paid overflow. The committed eligibility report contains reproduction commands and the prepared, unsent provider question.

    Exact owner-controlled needs: applicable provider support evidence for the public-target workflow, account-specific worker/paid-overflow evidence, and explicit authorization for a new selected-repository App identity experiment and live activation. No reversible fake can establish those external facts. Offline implementation proceeded under the merge-mission authority; API billing, private mirrors, a machine user and native R1-only substitution remain outside the requested default.

  3. HemSoft commented on Oct 19, 2026

    @HemSoft
    CollaboratorAuthor

    PR #183 is merged as 3cc932d. The one owner-authorized retry completed with3P2/noP1 at36fdb420; all three findings are fixed in579fb35 and all17threads resolved. Vet and the full race suite passed. The merge used the invoked bounded-review rule; corrected579fb35 was tested but not freshly clean-reviewed. No GitHub CI pass is claimed. Actions remain disabled and no live reviewer was activated.

    Remaining acceptance: Applicable public-target subscription support, account worker/overflow evidence and exact alpha App identity/activation authority.

    Attempt1: Reviewed the account-specific support/CI sources and general scripted/workspace-token alternatives in the committed eligibility report; these do not establish applicable support for this public target.

    Attempt2: Compared native bot schema and authenticated installation inventory; only existing connected Codex/Vercel apps are present, with no supplied alpha identity. Interactive client metadata establishes no worker credential or overflow decision.

    Exact next owner/external action: Supply applicable provider/account evidence and explicitly authorize the selected-repository alpha App and live experiment.

    Provider support and owner credentials/spending/activation are external; neither changing visibility nor installing an App is inferred. This issue remains open; offline source completion is not live acceptance.

  4. HemSoft commented on Oct 19, 2026

    @HemSoft
    CollaboratorAuthor

    Fresh read-only support revisit while PR #184 is under review. The non-interactive guide and cached-account CI guide still exclude public/open-source targets for that account-auth workflow. The access-token alternative documents Business/Enterprise workspace credentials, but supplies no applicable account grant, public-target exception or worker credential for this owner. These are two independently inspected routes; neither clears this issue. Authenticated inventory still identifies this repository as public and lists only existing Codex/Vercel installations. Exact need remains applicable provider/account evidence plus explicit alpha App/worker qualification authority. No provider contact, spending, account switch, new credential or installation was performed.

  5. HemSoft commented on Oct 19, 2026

    @HemSoft
    CollaboratorAuthor

    PR #184 merged as 67077bd. Exact head8c3bdb3 passed vet/race and a clean connected-Codex review with no findings. Issue #175 is closed after its explicit fake-first protocol/configuration contract was fully audited. The new source broker and actual fake-worker executable path address the prior read-capture implementation gap. This does not qualify official-client confinement or activate the alpha.

    Remaining: Applicable public-target subscription/account route, worker/overflow evidence and exact alpha App/activation authority.

    Attempt1: Freshly fetched non-interactive and cached-account CI guides. Both exclude public/open-source targets for that documented cached-account workflow. This repository remains public.

    Attempt2: Freshly fetched Business/Enterprise access-token documentation and authenticated App inventory. This alternative establishes no account grant or new alpha identity for this owner; inventory still has only existing Codex/Vercel.

    Exact next owner/external action: Supply an applicable supported account/workspace route and worker/overflow evidence, then explicitly authorize the selected-repository alpha App/live experiment.

    No provider-policy exception, spending, external credentials or App activation can be inferred. A private host alone establishes no exception. This issue stays open. Sources: non-interactive auth, cached-account CI, workspace access tokens.

  6. HemSoft commented on Oct 20, 2026

    @HemSoft
    CollaboratorAuthor

    Documentation update, October 20, 2026. OpenAI documents a separate subscription-backed option for open-source/self-hosted tools through SIWC, including an official Codex app-server integration and self-hosted VMs.

    This uses a separately registered OAuth client, explicit ChatGPT-plan consent, a custom Responses provider and app-managed token refresh. It is a candidate integration to qualify offline. It does not establish support for the existing cached auth_mode=chatgpt / codex exec route against this public repository. The cached-account CI guide expressly excludes public/open-source targets and distinguishes external-token integrations.

    Remaining evidence includes this polling service's exact deployment classification, compatibility with official Codex 0.163.0 and the required model/tools, actual account/workspace/model eligibility, disabled credit spending and any other paid fallback, and the changed authentication/process contract. Usage controls describe app credit settings but do not prove the owner's setting. This establishes nothing about Claude eligibility or native GitHub App picker support.

    No registration, auth profiles, credentials, account settings, inference, paid usage, App installation or deployment was touched. The issue remains open. Offline implementation and fake qualification continue under the existing freeze exception; these live gates remain separate.

  7. HemSoft commented on Oct 21, 2026

    @HemSoft
    CollaboratorAuthor

    Dated model-free report merged in #222. #221 is closed unmerged. New App identity, exact selected scope and COMMENT authorship passed; native UI/API request creation failed in the checked sessions. Dedicated post-comment GraphQL rerequestReviews was untested, so API impossibility is not claimed. Dependent native lifecycle gates were unexercised. Franz superseded native assignment at 01:19 EDT October21 with @sfl review plus automatic PR-open/head-push polling; issues #172/#174/#176/#177 reflect that decision. Offline trigger implementation is in progress. Provider/account, isolation/recovery, benchmark/promotion and explicit live-activation gates remain open; no model inference or alpha activation occurred.

  8. HemSoft commented on Oct 22, 2026

    @HemSoft
    CollaboratorAuthor

    Read-only official-documentation recheck, October 21, 2026 at 20:15 EDT:

    • Both OpenAI's advanced CI account-auth guide and non-interactive guide still exclude public/open-source repositories from the described cached-account CI workflow. The workspace access-token alternative is documented for Business/Enterprise; no applicable personal-account entitlement or exception was established. A private polling host remains an unresolved applicability question, not an assumed exception.
    • Claude's authentication page documents a one-year subscription OAuth token for scripts/CI and notes that bare mode does not read it. Its credential-use guidance distinguishes ordinary individual native-client use from services acting for other users. These mechanisms alone do not establish this exact account/workflow's eligibility.
    • Codex pricing distinguishes local review usage from native GitHub review usage and describes credit-based continuation. Claude usage-credit guidance documents paid continuation and disabling usage credits. Neither page proves Franz's actual settings, model access or enforceable paid-overflow prevention.

    Result: provider applicability and account/overflow gates remain unproved. No login, auth file, billing setting, model request or support-message send was used. PR #233 prepares the selected comment/automatic polling questions; sending them still requires owner authorization. The completed model-free App evidence remains unchanged and this issue stays open.

  9. HemSoft commented on Oct 23, 2026

    @HemSoft
    CollaboratorAuthor

    Status as of October 23, 2026 (US Eastern): blocked; needs Franz. Acceptance items 1 and 2 and all four Validation items are still unchecked. Nothing in this mission sent a provider question, logged in, changed billing or ran inference.

    Two self-serve attempts on October 22, between about 10:10 and 10:19 AM EDT:

    1. Documents. OpenAI's CI auth guide says "Do not use this workflow for public or open-source repositories" and makes no exception. Anthropic's legal and compliance page says subscription OAuth is "intended ... to support ordinary use". It says Pro/Max limits assume "ordinary, individual usage" and sends auth-method questions to sales. Neither page covers unattended, owner-only review of a public repository, either way.
    2. Search for any 2026 exception for public repositories or unattended use. None found. OpenAI's CI guidance recommends API keys for CI, and API billing is excluded here.

    No reversible default can grant provider permission, so this cannot pass without an owner action.

    New since the last update: #248 recorded live Codex use as blocked on a separate ground. Hostile model metadata makes the pinned client restore tools. Codex code mode also needs an owner decision (details). Even a favourable OpenAI answer would not unblock live Codex use by itself.

    Decisions needed from Franz:

    • Authorize sending the prepared provider questions to OpenAI and Anthropic, or record a different owner decision.
    • Supply account-specific evidence for each account: model entitlement, worker auth and refresh, and paid overflow (Codex credits, Claude usage credits) disabled or enforceably blocked.
    • Give explicit live-activation authority, separately, once the other gates pass.

    Two validation items may be out of date. Validation 1 (follow the model-free steps, inspect request and re-request IDs) and Validation 3 (picker and request screenshots) were written for the native picker. The October 21 trigger decision superseded the picker. These items need Franz to re-scope them to the @sfl review and automatic polling triggers, or mark them superseded. This draft does not decide that.

    Recommendation: keep open. Its criteria are not met, and the remaining gates are external or need the owner.

    🤖 Generated with Claude Code

  10. HemSoft commented on Oct 27, 2026

    @HemSoft
    CollaboratorAuthor

    Owner decision (Franz, Discord 2026-10-27 12:49 ET): "yes" to the Chief's brief question "SFL: allow subscription use (#173) …? Rec: yes". This authorizes the first decision listed above: send the prepared provider questions to OpenAI and Anthropic, and use the subscription for owner-only SFL review once the remaining gates pass. Unchanged, per this issue: account-specific evidence (entitlement, worker auth and refresh, paid overflow disabled), the #248 hostile-metadata and Codex code-mode block, and a separate explicit live-activation go. The Chief will dispatch the next owner when pacing allows. All subscriptions are over pace right now.

  11. HemSoft commented on Oct 27, 2026

    @HemSoft
    CollaboratorAuthor

    Prepared the authorized, unsent provider support pack in PR #283, head 7fea453f04551cf69c0aac65962c3fe779f8c8f8. The send pack gives verified OpenAI and Anthropic Product Support routes and complete copy-ready messages for the exact owner-only unattended public-PR workflow, pinned official clients/models, worker auth/refresh, entitlement and enforceable paid-overflow prevention.

    Franz must submit through his own accounts under this assignment's no-login instruction. This worker sent neither message and created no support case. The October 27 subscription-use approval remains conditional on the remaining gates, and live activation still needs a separate go. This issue stays open; no provider permission, account evidence, billing change or live use is claimed.

    Council unavailable under V1.4 pacing from Chief's live /v1/capacity report: Claude over pace, Muse quota guard unknown. No activation design or gate changed. The SFL Discord post ID was absent, so the milestone is included in the Chief handoff.

  12. HemSoft commented on Oct 28, 2026

    @HemSoft
    CollaboratorAuthor

    Park unattended CLI and custom-App questions. Verify only the selected native account, repository access and billing boundary.

    Only "verify the selected native account, repository access and billing boundary" remains active. The native-review plan uses connected Codex review rather than an unattended CLI service or custom App.

    The parked work stops. Revival requires the original unresolved gates, as stated in the approved plan.

  13. added
    parkedWork parked pending its documented revival condition
    on Oct 28, 2026
  14. HemSoft commented on Oct 28, 2026

    @HemSoft
    CollaboratorAuthor

    Franz's October 28, 00:47 EDT correction supersedes the native-Codex product framing. SFL v0 uses our one T3 review session after an owner @sfl review request and publishes through the existing SFL App.

    The remaining native-account/provider-question slice of #173 is not needed for v0. The unattended dual-CLI alpha and its unresolved gates remain parked. This does not close the issue, claim its historical gates passed, or authorize revival of the alpha.

    Corrected plan: #285
    Implementation: https://github.com/hemsoft-dev/project-manager/pull/63

  15. HemSoft commented on Oct 29, 2026

    @HemSoft
    CollaboratorAuthor

    Closed per Franz (2026-10-29 19:34 ET, set-it-free-loop post: 'Agreed'): SFL stays inside pmd (project-manager). The standalone alpha controller, adapters and publisher are not being built. Next SFL work is project-manager #75 (review status) and #76 (changes between rounds). Reopen if that direction changes.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requestparkedWork parked pending its documented revival condition

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions