Update agentic docs to match current repo state
Changes across 4 wiki pages:
Main page (Using-AI-to-Work-on-.NET-MAUI):
- Fix agents table: correct file paths (.agent.md suffix), add Agentic
Workflows Agent and MAUI Expert Reviewer, rename PR Agent to PR Review
Workflow with correct location (skill + script, not agent .md file)
- Rewrite skills section: remove nonexistent pr-build-status and
ai-summary-comment, add 9 missing skills (pr-review, code-review,
azdo-build-investigator, run-helix-tests, run-integration-tests,
evaluate-pr-tests, find-regression-risk, agentic-labeler, dependency-flow)
- Reorganize skill categories: Testing (7), PR Review (5), Triage (3),
CI & Infrastructure (3)
- Add new Automated Workflows section covering 3 gh-aw workflows
- Add new Instruction Files section with category overview
- Update Repository Structure tree to match actual .github/ layout
Agent-Skills page:
- Add 3 missing skills: find-regression-risk, agentic-labeler,
dependency-flow
PR-Agent page:
- Update model list: Claude Sonnet 4.6 → Claude Opus 4.7,
Gemini-3-Pro-Preview → GPT-5.5
Learn-From-PR-Agent page:
- Fix agent file path: learn-from-pr.md → learn-from-pr.agent.md
Add design philosophy section to PR Agent wiki
- Core question: 'How would AI solve this with zero human influence?'
- PR Review = Issue Fixing equivalence
- Generation (try-fix) vs hardening (report) distinction
- Feedback loop for improving try-fix over time
- Strategic direction toward AI-first fixes
- Updated try-fix and report descriptions for consistency
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Fix PR-Agent and Agent-Skills accuracy
PR-Agent:
- Gate runs verify-tests-fail.ps1 directly (not a Copilot agent)
- Gate does NOT create tests — recommends @copilot write tests if none found
- Step 2 invokes Copilot CLI which triggers pr-review skill (not code-review)
- Remove code-review from skills used by pipeline (it's standalone only)
- Add Review-PR.ps1 CLI usage examples
- Fix output directory structure to match actual paths
- Add EstablishBrokenBaseline.ps1 -Restore between attempts
Agent-Skills:
- Mark code-review as standalone (not part of automated pipeline)
- Remove non-existent pr-build-status
- Clarify try-fix is invoked by pr-review, not standalone
- Note evaluate-pr-tests also has gh-aw /evaluate-tests trigger
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Update PR-Agent and Agent-Skills wiki to match current architecture
- PR-Agent: Fix phase descriptions to match Review-PR.ps1 (4 steps, not 5 phases)
- PR-Agent: Update models to current set (Opus 4.6, Sonnet 4.6, GPT-5.3-Codex, Gemini-3-Pro)
- PR-Agent: Fix file locations (Review-PR.ps1, not .github/agents/pr.md)
- PR-Agent: Add cross-pollination, agent label system, state tracking docs
- Agent-Skills: Add missing skills (code-review, pr-finalize, evaluate-pr-tests, azdo-build-investigator)
- Agent-Skills: Remove non-existent pr-build-status skill
- Agent-Skills: Organize into Testing, PR Review, Learning, Infrastructure categories
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Updated PR Agent (markdown)
Updated PR Agent (markdown)
Updated PR Agent (markdown)
Updated PR Agent (markdown)
Updated PR Agent (markdown)
Updated PR Agent (markdown)
Updated PR Agent (markdown)
Updated PR Reviewer Agent (markdown)