[seanspeaks]

ADR-014 and its amendment settle how an ADR is filed. Nothing settles how one is decided. This proposes that.

What the register currently shows

Measured on next — 28 merged ADRs, 14 more unmerged across nine open PRs:

• 15 of 28 merged ADRs (54%) are still Proposed. ADR-015→026 are all dated 2026-06-09 and haven't moved in three months. ADR-004 has been Proposed since 2025-01-25 — twenty months.
• No ADR PR has ever had a human approval. reviewDecision is empty on all nine. Every review on them is a bot (graphite-app, cursor, claude, chatgpt-codex-connector, github-advanced-security) — all COMMENTED, none APPROVED.
• The vocabulary is already violated. The README defines four statuses; ADR-018 is Implemented, which isn't one of them.
• The README points at a process that doesn't exist. It describes Proposed as "Under discussion (use RFCs for new proposals)". There is no RFC directory, template or process anywhere in this repo.
• ADR-031 is Proposed while fix(core): single-flight the 401 token refresh in Requester #636 implements half of it. Two merged ADRs carry Deciders: TBD.

Proposed has stopped carrying information. It means "merged" — because merging is the only event anyone performs, and it changes no status.

Why not just add RFCs

A two-tier split (RFC explores, ADR records) is the standard recommendation, and I'm proposing we decline it — for reasons specific to this repo:

1. ADR-014 just merged two documentation bodies into one, and the reconciliation after it had to resolve eight number collisions across six branches. A second register isn't a theoretical cost; we've just paid it once.
2. The RFC's job is already done by a draft PR, with review threads anchored to the actual text. docs(adr): three new ADRs — ADR-028 App Init, ADR-029 Skills, ADR-030 Agent Pipeline #619 and docs(adr): draft ADR-030 Integration Versioning (exploratory) #620 already work exactly this way.
3. Our constraint is deciding, not authoring. Two named deciders, fourteen unreviewed ADRs. A second tier adds authoring — the part that is already cheap — and does nothing for review, which is the part that is scarce.
4. Agent readers make one authoritative register worth more, not less. Splitting into "proposals here, decisions there" doubles the ambiguity at exactly the moment ADR-021 (Ontology) and ADR-036 (Skills) point agents at it.

The gap isn't a missing proposal stage. It's that nothing marks the moment a proposal becomes a decision.

The proposal

Merging is the decision event. An ADR does not merge as Proposed.

If it isn't ready to be accepted it stays a draft PR — still numbered, still reviewable, still linkable, just not yet in the register as a decision. If it's genuinely undecidable today it merges as Exploratory with its open question stated.

transition	performed by
Proposed → Accepted	merging the ADR PR, with an approving review from a named decider
Accepted → Implemented	the code PR that ships it, in the same commit as the code
Accepted → Superseded	the superseding ADR, which must name its predecessor

Every transition rides along with work someone is already doing. Nothing to remember, nothing to schedule.

Also: adds Exploratory (we have a real one — Integration Versioning, whose own status line says the decision is deliberately deferred), legitimises Implemented (already in use on ADR-018), requires named deciders, and extends the existing adr-number-check workflow to enforce all of it.

Note on its own status

This is marked Proposed, which under its own §3 means it cannot merge that way. That's intentional: merging it means flipping it to Accepted in the same PR. If it isn't accepted, it shouldn't land.

Consequences worth arguing with

• A named approval becomes required to merge an ADR. With two deciders that's a real constraint, and the current fourteen have to be worked through rather than bulk-merged.
• The 15 merged Proposed ADRs need a one-time pass to set real statuses — a judgement call per ADR, not automatable.
• Draft PRs become load-bearing; the PR list becomes the backlog view.
• Implemented depends on someone flipping it in the code PR. Skipped, the ADR understates reality — softer than today's failure, but still one.

Sequencing

Independent of #644 and #636. 043 is the next free number after #644's 041 and #636's 042, so all three can land in any order.

🤖 Generated with Claude Code

[claude]

⚠️ Code review skipped — your organization's overage spend limit has been reached.

Code review is billed via overage credits. To resume reviews, an organization admin can raise the monthly limit at claude.ai/admin-settings/claude-code.

Once credits are available, comment @claude review on this pull request, or push a new commit, to trigger a review.

[netlify]

✅ Deploy Preview for friggframework-org ready!

Name	Link
🔨 Latest commit	deecd71
🔍 Latest deploy log	https://app.netlify.com/projects/friggframework-org/deploys/6aa5f02639f94100088170cb
😎 Deploy Preview	https://deploy-preview-646--friggframework-org.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[seanspeaks]

ADR reconciliation needed after Stack A/B — drafted with Claude Code at Sean's request

Severity: cross-ref only. The new stacks test ADR-043's status rules in three ways, and it may be worth naming each one:

• Status flipped by a code PR. ADR-063 is Proposed in its ADR PR (docs(adr): ADR-063 integration definitions as data #692), and code PR feat(core): published definitions are integration types like code; code can embed one; workspaces own their own #696 changes it to Accepted (2026-10-19) while docs(adr): ADR-063 integration definitions as data #692 is unmerged. §2 says Accepted = the ADR PR merges and Implemented = the code PR. Add: "a code PR never moves an unmerged ADR's status".
• Amendment files. feat(core): several named proxy base URLs per module (F8) #690 adds 052-amendment-1-named-proxy-bases.md marked Accepted for ADR-052, which is itself unmerged (docs(adr): ADR-052 entity proxy security, ADR-053 Management API versioning #668). Decide whether amendments are separate files or edits in place. Check that the number-claim CI (ADR-014 amendment) does not read a second 052- file as a second claim.
• Withdrawn numbers. docs(adr): ADR-056 §3.1–3.6 ownership overrides + multi-org membership; ADR-059 app-level entity registry #677 folds the drafted ADR-058 into ADR-056, so 058 has no file. ADR-063 cites "ADR-062 (proposed, not yet written)". Add a Withdrawn/Reserved convention (a register row with no file), so CI and readers know 058 must not be reused and 062 is held. 060 and 061 are free. docs(adr): reconcile the register — 13 unmerged ADRs into one contiguous sequence #644 proposes 060 for the stray marketplace ADR.