retrofit: mint infrastructure-helpers capability (3 REQs / 6 methods)

[rubenvdlinde]

Retrofit — Reverse-Spec

Mints a new `infrastructure-helpers` capability covering 6 methods across 3 cross-cutting helper classes flagged Bucket 2b by the 2026-05-24 coverage scan (no owning capability).

Ghost change: `archive/2026-05-24-retrofit-infrastructure-helpers` (archived in this PR).

What this PR does

• Creates a new `openspec/specs/infrastructure-helpers/spec.md` with `retrofit: true` frontmatter
• Drafts 3 REQs (one per helper class):
  • REQ-INFRA-001 — SlugGenerator: slug grammar, transform pipeline, validation
  • REQ-INFRA-002 — UserAttributeResolver: 4 supported attributes + 5 comparison operators
  • REQ-INFRA-003 — RequestDataExtractor: tile + placement form data extraction
• Creates tasks.md with one task per REQ (all `[x]` — code already exists)
• Annotates 6 methods + 3 files with `@spec openspec/changes/archive/2026-05-24-retrofit-infrastructure-helpers/tasks.md#task-N` (Phase 2's existing `@spec` on `UserAttributeResolver::getUserAttributeValue` is preserved; new tag is appended)
• Archives the change

Why a new capability rather than `--extend`?

Per ADR-003 and the SKILL's bias-toward-`--extend` rule, considered alternatives:

Option	Why rejected
`--extend dashboards` (SlugGenerator)	Slug helper is reusable beyond dashboards
`--extend conditional-visibility` (UserAttributeResolver)	Resolver is a generic primitive; the visibility logic that uses it stays put
Per-method splits across `tiles` + `widgets` (RequestDataExtractor)	Would duplicate the helper's contract across two specs
Split into three different capabilities	Would scatter the helpers' shared architectural shape

The new capability includes a `## Scope` section that prevents drift: future helpers that DO fit a domain capability should land there, not here.

What this PR does NOT do

• No code behaviour changes — pure annotation + spec
• Does not silently fix observed-but-suspicious behaviour:
  • `RequestDataExtractor::extractPlacementData()` type-casting asymmetry vs `extractTileData()` — documented in spec Notes
  • `UserAttributeResolver::evaluateOperator()` silent-false on unknown operators — documented (future-tightening TODO)
  • `SlugGenerator::slugify('')` empty-string return — documented as intentional (callers substitute UUID)

Review focus

• REQ language matches observed behaviour (not aspirational)
• Scenarios are testable
• Capability scope is well-defined enough to prevent drift
• New capability genuinely earns its slot (not just a dumping ground)

Source: `openspec/coverage-report.md` generated 2026-05-24
Cluster: infrastructure-helpers (Bucket 2b)
Refs #292

[github-actions]

Quality Report — ConductionNL/mydash @ 605274c

Check	PHP	Vue	Security	License	Tests
lint	❌
phpcs	❌
phpmd	❌
psalm	❌
phpstan	❌
phpmetrics	❌
eslint		❌
stylelint		❌
composer			❌	❌
npm			✅	✅ 436/436
PHPUnit					⏭️
Newman					⏭️
Playwright					⏭️

---

Quality workflow — 2026-05-24 18:23 UTC

Download the full PDF report from the workflow artifacts.