Repository navigation
docs(plan): record user decisions E2, E7, E8 and SETUP-14 progress - #54
Conversation
E2: W4-01 benchmark shrinks to 5 tasks. E7: extended W5 core confirmed. E8: Codex gpt-6-astra first for seams/security confirmed. SETUP-14: ollama sign-in and OpenCode models observed working on 2026-10-02. No-Test: documentation-only plan status update; no runtime behavior changed. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
Copilot review overview
🟡 Changes recommended
The generated Dev HQ snapshots must be updated to reflect the plan changes.
Review effort: Balanced
Findings: 1
Open (1)
What changed in this PR
Records user decisions and SETUP-14 progress in the project plan.
Changes:
- Confirms E2, E7, and E8.
- Fixes W4-01 at five benchmark tasks.
- Marks SETUP-14 partially complete.
| File | Description |
|---|---|
docs/PLAN.md |
Updates decisions, benchmark scope, and setup progress. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
| | SETUP-09 | Lokaler Review-Lauf `scripts/review/run-local.sh` | S | doc | ✓ #39 | | ||
| | SETUP-12 | Rest des Docs-only-Pfadfilters, soweit CI-02/CI-03 ihn nicht abdecken | S | ci | offen | | ||
| | SETUP-14 | Nutzer: tote Keys, OpenCode-Modelle, `ollama signin`, Permission-Regeln | S | N | offen | | ||
| | SETUP-14 | Nutzer: tote Keys, OpenCode-Modelle, `ollama signin`, Permission-Regeln | S | N | teilweise: `ollama signin` und OpenCode-Modelle geprüft (02.10.); offen: tote Keys, Permission-Regeln | |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 25914a85a3
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".
| | SETUP-09 | Lokaler Review-Lauf `scripts/review/run-local.sh` | S | doc | ✓ #39 | | ||
| | SETUP-12 | Rest des Docs-only-Pfadfilters, soweit CI-02/CI-03 ihn nicht abdecken | S | ci | offen | | ||
| | SETUP-14 | Nutzer: tote Keys, OpenCode-Modelle, `ollama signin`, Permission-Regeln | S | N | offen | | ||
| | SETUP-14 | Nutzer: tote Keys, OpenCode-Modelle, `ollama signin`, Permission-Regeln | S | N | teilweise: `ollama signin` und OpenCode-Modelle geprüft (02.10.); offen: tote Keys, Permission-Regeln | |
There was a problem hiding this comment.
When npm run hq:live regenerates the DevHQ snapshot, milestoneState() in scripts/lib/hq-parse.mjs classifies any Stand value without ✓ as open; therefore this new teilweise: value is displayed as fully open rather than in progress. Express the completed items using the existing mixed ✓ …; offen … vocabulary, or extend the parser and its tests to recognize teilweise, so DevHQ accurately reflects the recorded progress.
Useful? React with 👍 / 👎.
| | SETUP-09 | Lokaler Review-Lauf `scripts/review/run-local.sh` | S | doc | ✓ #39 | | ||
| | SETUP-12 | Rest des Docs-only-Pfadfilters, soweit CI-02/CI-03 ihn nicht abdecken | S | ci | offen | | ||
| | SETUP-14 | Nutzer: tote Keys, OpenCode-Modelle, `ollama signin`, Permission-Regeln | S | N | offen | | ||
| | SETUP-14 | Nutzer: tote Keys, OpenCode-Modelle, `ollama signin`, Permission-Regeln | S | N | teilweise: `ollama signin` und OpenCode-Modelle geprüft (02.10.); offen: tote Keys, Permission-Regeln | |
There was a problem hiding this comment.
Marking the OpenCode models as checked here leaves docs/setup/opencode.md:22-24 saying that reachability of opencode-go/glm-5.3-flash is still unproven on this PC. Because provider setup and its observed state live under docs/setup/, agents following that page will continue treating this as an outstanding user task and may repeat or block setup work; update the setup page with the observed 02.10 command and result alongside this status change.
AGENTS.md reference: AGENTS.md:L68-L70
Useful? React with 👍 / 👎.
| | E7 | W5-Kern: beschlossen waren W5-22, W5-28, W5-02a und Not-Aus; PLAN-01 hat zusätzlich W5-00b, W5-02b3–b5/b7 (Report-Folgearbeiten aus den W5-02-Reviews) und den Prüfpfad W5-05 in M3/M4 eingeordnet | ja, erweiterten Kern bestätigen (Review PR #175, kimi-k3 F-3) | offen (Nutzer) | | ||
| | E8 | Routing Nahtstellen/Security: `docs/setup/providers.md` routet primär auf Codex `gpt-6-astra`, Claude-Worker nur als Ausweichen — die alte Modellregel (Claude implementiert Nahtstellen/Security) ist damit ersetzt | ja, Routing bestätigen (Review PR #175, kimi-k3 F-4) | offen (Nutzer) | | ||
| | E7 | W5-Kern: beschlossen waren W5-22, W5-28, W5-02a und Not-Aus; PLAN-01 hat zusätzlich W5-00b, W5-02b3–b5/b7 (Report-Folgearbeiten aus den W5-02-Reviews) und den Prüfpfad W5-05 in M3/M4 eingeordnet | ja, erweiterten Kern bestätigen (Review PR #175, kimi-k3 F-3) | ✓ bestätigt (Nutzer 02.10.) | | ||
| | E8 | Routing Nahtstellen/Security: `docs/setup/providers.md` routet primär auf Codex `gpt-6-astra`, Claude-Worker nur als Ausweichen — die alte Modellregel (Claude implementiert Nahtstellen/Security) ist damit ersetzt | ja, Routing bestätigen (Review PR #175, kimi-k3 F-4) | ✓ bestätigt (Nutzer 02.10.) | |
There was a problem hiding this comment.
Confirming Codex as the primary security route here leaves docs/setup/codex.md:60-62 instructing agents to keep security lanes on Claude because GPT-6 Astra has not been re-probed after earlier security-task refusals. Since provider-specific operating guidance lives under docs/setup/, the two active instructions now select different workers for the same package; update that page to reflect the confirmed route or retain an explicit exception until the required probe exists.
AGENTS.md reference: AGENTS.md:L68-L70
Useful? React with 👍 / 👎.
|
Queued — the merge queue status continues in this comment ↓. |
|
@Mergifyio queue |
Merge Queue Status
This pull request spent 12 minutes 16 seconds in the queue, including 11 minutes 17 seconds running CI. Required conditions to merge
|
Deine Entscheidungen vom 02.10. stehen jetzt im Plan: Der Leistungstest bekommt 5 Aufgaben, der erweiterte Sicherheits-Kern ist bestätigt, und heikle Stellen baut zuerst Codex. Bei SETUP-14 sind Ollama-Anmeldung und OpenCode-Modelle geprüft. Offen bleiben dort nur die alten Schlüssel und die Berechtigungsregeln.
Report
What changed (
docs/PLAN.mdonly)ollama signinand OpenCode models observed working.Evidence
ollama run glm-5.2:cloud "Antworte nur mit: ok"→ok, exit 0 (2026-10-02).opencode modelslists theopencode-go/*models, exit 0.git commitwith the precommit lane → all gates green (secrets, fmt, cargo-check, typecheck), commit25914a8.NICHT ABGEDECKT
prepushlane not run locally (docs only); pushed withPA_PREPUSH=light.#[cfg(unix)]tests do not compile on Windows.Review disposition
Tier C (documentation): no external review, the gates suffice.
🤖 Generated with Claude Code