ADO Integration

Structured markdown comments posted to stories and bugs. Each skill uses a unique signature tag (e.g., [DoRAgent], [DevPlan]) for idempotent updates — re-running a skill updates the existing comment instead of creating duplicates.

Creates child work items under stories: Task items for sprint planning and Bug items from QA verification. All created items are linked as children of the parent story.

Technical documentation and PR review reports published to the ADO Wiki under sprint-organized paths. Generated from spec files, retroactively from PRs, or from code review findings. Also supports Confluence.

Definition of Ready checklist with interactive checkboxes. BAs can check items directly in ADO, then re-run DoR to update the scorecard. Three modes: full post, update, or skip if unchanged.

Short development plan for the team: scope (S/M/L) and 3-5 bullets on what will change. Suitable for pasting into Teams. Posted before development starts — no wiki link (wiki is generated post-implementation).

Definition of Done validation results. Checks PR status, approving reviews, resolved threads, child tasks, tests, build, secrets, and accessibility. Posts pass/fail verdict with score.

Requirements Engineering summary. Structured requirements spec with task breakdown, identified risks, and effort estimates. Posted after the RE agent analyzes the story.

Estimation results: hours per task, story points, affected AEM pages, and complexity breakdown. Uses an HTML comment signature for machine-readable detection.

Clarification questions posted when a bug description is ambiguous. Asks for missing reproduction steps, expected behavior, or environment details to unblock investigation.

Bug verification evidence. Before mode: confirms the bug is reproducible on QA (screenshots, console errors). After mode: confirms the fix works on local AEM. QA mode: confirms the fix is deployed and working on QA post-merge.

Short QA handoff pointer: QA page URLs (author + publisher), prerequisites, what changed (3-5 bullets), and a link to the wiki page for full details. QA has test plans for detailed verification — the comment is a notification, not a test script.

Creates Bug work items when QA verification finds UI/functional issues on AEM. Each bug includes: repro steps, expected vs actual behavior, severity (High/Medium), screenshot evidence, and page URL. Always asks for human confirmation before creating.

The Requirements Engineering agent creates child Task work items under the parent story. Each task has a title, description, type classification (FE/BE/Authoring), and is linked as a child of the story.

Batch-creates child Task work items with hour estimates. Sets Original Estimate and Remaining Work fields. With close argument: moves Remaining Work to Completed Work, zeros remaining, and sets state to Closed. Full task lifecycle: create → close.

Creates Task work items for failed Definition of Done criteria that cannot be auto-fixed. Auto-fixes what it can (tests, docs, changelog), then creates ADO Tasks for the rest.

Generates wiki page structured as a demo walkthrough: Summary, Design Reference (Figma), What Changed and Why, QA Verification (URLs), Dialog Changes (screenshot), Frontend Preview (screenshot), Figma Comparison (side-by-side), Authoring Guide, Files Changed. Screenshots use repo-relative paths for committing. Posts to ADO Wiki under <wiki-doc-root>/<Sprint>/<id>-<slug>.

Generates documentation retroactively for stories that shipped without docs. Fetches the ADO story, finds linked PRs, searches the codebase for changes, and produces wiki-ready documentation. No spec files needed.

Generates a categorized report from a single PR review — groups comments by category (accessibility, bug, functionality, regression), tracks patch resolution, and posts to ADO Wiki under <wiki-pr-review-root>/<ticket>.

Generates categorized reports for multiple PR reviews in batch — lists reviewed PRs, generates a report for each, and posts all to ADO Wiki. Ideal for sprint review summaries and tech lead dashboards.

After creating child Task items, updates their estimation fields: Original Estimate and Remaining Work. Uses batch update for efficiency. Also sets assignment if configured.

Closes all open child tasks after development: moves Remaining Work into Completed Work, zeros remaining, sets state to Closed. Checks implement.md to verify all steps are done before closing. Sprint burndown and capacity stay accurate.