Files
breadcrumb-the-shire/.agents
fs 53ccd212d5 docs(agents): encode list-export convention as GR-UI-EXPORT guard
Make the core-primitive-end-to-end rule for Grid.js list exports binding
instead of advisory — so future consumers (and reviewers) cannot quietly
drift back into inline fputcsv / hand-rolled headers / bespoke click
listeners / lurl() for query-carrying URLs.

- CLAUDE.md: new UI Patterns bullet + two "Never Do This" entries
  spelling out the server/view/client contract in one place.
- .agents/checks/guard-catalog.json: new GR-UI-EXPORT entry describing
  the end-to-end requirement (exportRequireGetRequest + exportCapLimit +
  exportSendCsv + ExportColumn + shared filter-schema + dropdown partial
  + endpointUrl() + initListExport).
- .agents/checks/guard-enforcement-map.json: wire the new guard to
  tests/Architecture/ListExportContractTest.php as the automated
  evidence source.
- .agents/checks/guard-checklist.md & .agents/prompts/reviewer-code.md:
  add GR-UI-EXPORT so the Code Reviewer prompt and the checklist flag
  violations alongside GR-UI-LIST.
- tests/Architecture/ListExportContractFiles: registry of every list
  export (currently helpdesk-domains, admin/users, audit/system-audit).
  Adding a new list = one entry, contract test covers the rest.
- tests/Architecture/ListExportContractTest: six checks per registered
  entry — endpoint uses the primitive, no hand-rolled output, data and
  export share the same filter-schema, template includes the dropdown
  partial and uses endpointUrl(), page module imports and invokes
  initListExport, and the shared dropdown partial stays zero-config.
- pages/admin/users/export().php: align with the new contract by
  switching from ad-hoc requestInput()->queryAll() reads to
  gridParseFiltersFromSchemaFile(__DIR__ . '/filter-schema.php'), same
  as the data endpoint — eliminates the last place Grid and export
  could disagree on what "the current filter" means.

Gates: PHPUnit 1900 OK (+6 contract checks), PHPStan 0 errors.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-21 22:50:43 +02:00
..

.agents/

Centralized home for the 7-role agent workflow system.

Structure

workflow.md         ← Roles, state machine, handover rules (start here)
contracts/          ← JSON schemas for role outputs (7 roles)
templates/          ← Starter JSON payloads per role (7 templates)
prompts/            ← Role-specific prompt baselines (7 prompts)
checks/             ← Guard catalog (27 guards), quality gates (9), checklists
                   ← Enforcement policy + guard enforcement map
skills/             ← Reusable skill packages (guardrails, planner, grid, PHP CI)
runs/               ← Runtime artifacts per workflow run (gitignored)

Roles

# Role Artifact Guards
1 Analyst analysis.json
2 Planner plan.json selects guards
3 Executor execution-report.json provides evidence
4 Code Reviewer review-code.json 17 guards (GR-CORE/TEST/UI)
5 Security Reviewer review-security.json 10 guards (GR-SEC)
6 Acceptance Tester review-acceptance.json SC-* criteria
7 Finalizer finalize.json all reviews + CI
What Where
Workflow & roles workflow.md
Guard catalog (27 guards) checks/guard-catalog.json
Quality gates (9 gates) checks/quality-gates.json
Enforcement policy checks/enforcement-policy.json
Guard enforcement map checks/guard-enforcement-map.json
Code review checklist checks/guard-checklist.md
Acceptance checklist checks/acceptance-checklist.md
Module manifest schema contracts/module-manifest.schema.json

Project Context

See CLAUDE.md (repo root) for tech stack, architecture rules, and conventions.