Add relative scheduler anchors (codex)

[numnx]

🧩 Task: Add relative scheduler anchors

Automated task PR opened by Code UX for T01.

📋 Summary

Sprint: Sprint 145 · Micro Improvements · Task: T01

Deliver SPR-145 micro improvements: Jira import In Work transitions, relative sprint scheduling from scheduler/composer/quicksprints, onboarding menu ordering, task-card quick actions, and sticky settings context controls.

📝 Full Task Prompt

## Objective
Add a first-class scheduler contract for entries that start when another sprint ends, while preserving existing absolute date/time scheduling and recurrence behavior.

## Scope
- `src/contracts/scheduler-types.ts`
- `src/repositories/scheduler-repository.ts`
- `src/services/scheduler-service.ts`
- `src/server/scheduler-routes.ts`
- `src/mcp/management/scheduler-actions.ts`
- `src/contracts/mcp-tool-definitions.ts`
- `tests/backend/mcp/management-scheduler-actions.test.ts`
- Add or update focused scheduler service/repository tests under `tests/backend/services/` or `tests/backend/repositories/`
- `docs/dashboard/scheduler.md`

## Implementation Requirements
1. Extend the scheduler contract with an optional schedule anchor for `after_sprint_end`, including the source sprint id and an optional non-negative offset in minutes; keep the existing absolute `scheduledFor` path as the default.
2. Persist and hydrate the new anchor without breaking existing `scheduler_entries` rows; if no migration is required because the anchor can live in the target payload, document that choice in code/tests.
3. Validate that the anchor sprint exists, belongs to the same project, and is not the same sprint being scheduled when the scheduled target is also a sprint.
4. Make due-entry execution treat anchored entries as due only after the anchor sprint is terminal (`completed`, `failed`, or `cancelled`) and the optional offset has elapsed from the sprint end timestamp (`endDate` or the latest terminal run `finishedAt` when available).
5. Ensure anchored one-time entries execute once and then complete, and reject recurrence for `after_sprint_end` unless the existing scheduler model can support it cleanly without ambiguous repeated runs.
6. Add MCP `manage_scheduler` fields for the anchor mode/source sprint id/offset and map them into `CreateSchedulerEntryInput` and `UpdateSchedulerEntryInput`.
7. Update scheduler documentation with the new backend contract, validation, and runtime semantics.

## Constraints
- Do not change behavior for existing absolute-time scheduler entries, memory-remediation settings schedules, or recurrence expansion.
- Do not require a destructive database migration for existing entries.
- Keep project isolation strict: an entry must not be schedulable from a sprint in another project.
- Avoid polling all historical data in hot paths; use repository helpers or bounded scans for scheduled anchored entries.

## Verification
- Run `pnpm run test:backend -- tests/backend/mcp/management-scheduler-actions.test.ts`.
- Run the new or updated focused scheduler service/repository tests.
- Run `pnpm run lint`.
- Success means absolute schedules still pass existing tests, anchored schedules only become due after the source sprint ends, and invalid cross-project/self-anchor inputs are rejected.

✅ QA Review

⏳ QA review runs after this PR is opened. Check the task card in the dashboard for the outcome once it lands.

🤖 Model & Provider

Provider	Codex CLI
Model	gpt-5.5

⏱️ Timing

Started	2026-07-06 21:49:56 UTC
Finished	2026-07-06 22:03:02 UTC
Duration	13m 5s

📊 Token Usage

Metric	Value
Input tokens	190,877
Cached input tokens	2,772,736
Output tokens	17,986
Total tokens	2,981,599
Invocations	1
Tool calls	62
Active time	8m 40s

Included usage estimate (subscription/local login): $2.88

🌿 Branch Info

Base: feature/CODUX-145-micro-improvements
Head: task/feature-codux-145-micro-improvements-t01-codex-mr9r6upa

---

🤖 Generated by Code UX