aggregator(alerts): bound mergeClasses keys to a class registry (#3702 disposition 4)

Disposition 4 of design #3702 Phase 3 (audit task 5613, 2026-11-03). mergeClasses (pm-stack/aggregator/src/alerts.ts:119) copies every boolean key from client input into the subscription's classes JSON, so a client can grow that object without bound; classEnabled treats an unknown key as on, so unknown keys are semantically inert today (test pm-stack/aggregator/test/alerts.test.ts:53 pins pass-through for quota-wall).

Accepted for v1 in the audit: no server-side amplification, one row per subscription, growth is client-cooperative, and #3317 is about to introduce real classes — a strict allowlist now would drop them. This issue is the bound once a class registry exists: cap the key count/length and validate against the registry of classes the engine actually consults.

Filed from #3702 Phase 3 audit; no behaviour change landed there by design.

Triage — live, split: cap lands now, registry waits for #3317

Verified at origin/main 048da81f1 (2026-11-03).

Confirmed: mergeClasses (pm-stack/aggregator/src/alerts.ts:119-129) copies every boolean key from client input with no count or key-length bound. classEnabled (:161-163) still treats unknown keys as on, so today's unknown keys are inert — the exposure is unbounded object growth inside one subscription row, exactly as the audit described. The audit's reasoning for accepting it in v1 was #3317.
feat(aggregator): notification engine — class triggers, dedupe, quiet hours, escalation (#1254 slice c) #3317 is still open (notification engine — class triggers/dedupe/quiet hours), so there is no registry to validate against yet. A strict allowlist now would drop the classes feat(aggregator): notification engine — class triggers, dedupe, quiet hours, escalation (#1254 slice c) #3317 adds.
Ordering: the size cap does not depend on the registry and does not change accepted-key semantics, so it is routed now; registry validation stays on this issue and lands with/after feat(aggregator): notification engine — class triggers, dedupe, quiet hours, escalation (#1254 slice c) #3317.
routed: queue task 8572 (repo-health:v1:fix:developing-today/code:issue:4981, kind small_fix, low risk, 45 min) — count + key-length cap in mergeClasses, the quota-wall pass-through test preserved, targeted tests for both caps. The issue's own recommendation ("cap the key count/length and validate against the registry") is satisfied in that order rather than held hostage to the registry.

🤖 Agent note · handed off

Agent: fleet monitor (repo-fleet-monitor)
Session / ID: task 8572 · repo-fleet-monitor run 2026-11-03 18:12 CST
Found via: monitor reconciliation
Task: Claim evaporated — task 8572 returned to queued and unclaimed in repo-health-code-pilot after the holding run ended without a terminal verb (checkpoint stage: handed_off). Labels reset to the truth: queue/queued + status/needs-triage. The task cap was also raised this run where it had sat dead-in-queue, so it is claimable again.
Expectation: next claimant picks up from the checkpoint stage named above.

🤖 Agent note · handed off

Agent: queue worker (repo-worker-code), lane code-h
Session / ID: task 8572
Found via: queue claim
Task: Small fix for issue aggregator(alerts): bound mergeClasses keys to a class registry (#3702 disposition 4) #4981 — re-verified fresh: issue aggregator(alerts): bound mergeClasses keys to a class registry (#3702 disposition 4) #4981 OPEN, no hard-stop labels; claim labels synced this shift (queue/claimed, work/assigned, status/in-progress on; queue/queued, status/needs-triage off). PR fix(aggregator): bound mergeClasses class keys by count and length (#4981) #5163 OPEN with no labels. Execution lane assigned opencode; handing off to the laptop claimant, not working directly. Note: task 8572 is at attempt 8/8 (final).
Expectation: claimant launches the bound opencode session to drive task 8572 to a terminal state.
Links: issue aggregator(alerts): bound mergeClasses keys to a class registry (#3702 disposition 4) #4981, PR fix(aggregator): bound mergeClasses class keys by count and length (#4981) #5163

agent-note:v1
state: picked_up
agent: repo-worker-code
lane: 1a180ae5-384f-4e71-8d77-08f7b1652c75
task: 8572
found_via: OpenChamber scheduled-lane queue claim
revalidation: issue #4981 OPEN; linked PR #5163 OPEN, head d607022
next: recheck outcome recorded; the PR is authored by dezren39 and the worker runbook forbids touching Drew-owned PRs.

agent-note:v1
state: blocked
agent: repo-worker-code
lane: 1a180ae5-384f-4e71-8d77-08f7b1652c75
task: 8572
reason_class: escalated-to-user
blocker: Owner action required: PR #5163 is authored by dezren39, and the worker protocol forbids touching or merging Drew-owned PRs. At head d607022 the only check is CodeRabbit neutral/rate-limited; there are no Actions runs or commit statuses, so the basic PR-tier check remains absent. Owner must take over or explicitly reassign/authorize a non-Drew owner, then requeue; this lane will not edit or merge the PR.
links: issue #4981; PR #5163

Navigation Menu

Repository navigation

aggregator(alerts): bound mergeClasses keys to a class registry (#3702 disposition 4) #4981

Triage — live, split: cap lands now, registry waits for #3317

28 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

aggregator(alerts): bound mergeClasses keys to a class registry (#3702 disposition 4) #4981

Description

Activity

dezren39 commented on Nov 3, 2026

Triage — live, split: cap lands now, registry waits for #3317

28 remaining items

dezren39 commented on Nov 4, 2026

dezren39 commented on Nov 4, 2026

dezren39 commented on Nov 4, 2026

dezren39 commented on Nov 4, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions