Command Quick Reference

All slash commands at a glance -- grouped by workflow phase

Setup

First-Time Setup

Run once per repo to configure the project.

CommandDescription
/dx-initConfigure project — ADO mapping, build commands, base branch. Creates .ai/config.yaml.
/aem-initDetect AEM structure, append aem: section to config. AEM projects only.
/dx-adaptAuto-detect project type, role, toolchain, and build commands. Writes project.type, project.role, and toolchain: section to config.
/dx-scanDeep project scan — discover FE/BE conventions from code, customize rules with real values, generate seed data. Interactive.
/auto-initScaffold autonomous agent infrastructure (optional).
Recon

Reconnaissance

Quick lookups before starting work.

CommandArgumentDescription
/dx-ticket-analyze<id>Quick ticket recon — files, pages, acceptance criteria
/aem-component<name>Find all source files + AEM pages for a component
/aem-page-search<name>Find all AEM pages using a specific component
/dx-help<question>Answer architecture questions from local docs
/dx-estimate<id>Analyze story, produce hours/SP estimation
/dx-council<question>LLM Council — 3 advisors analyze, peer-review, synthesize verdict. Use council the plan or council the implementation.
Phase 1

Requirements

Fetch, validate, explain, research, and share.

CommandArgumentDescriptionOutput
/dx-req<id>Full requirements pipeline (fetch, DoR, explain, research, share). Detects cross-repo scope when repos: exists in config.All spec files + ADO comments
/dx-req-tasks<id>Create child Task work items with estimatesADO/Jira tasks
/dx-req-import<path>Validate external requirements documentexplain.md
Figma

Figma Design-to-Code

Extract, prototype, and verify from Figma designs.

CommandArgumentDescription
/dx-figma-all[id] [URL]Full Figma workflow: extract, prototype, verify
/dx-figma-extract[id] [URL]Extract design context, tokens, screenshots from Figma
/dx-figma-prototype<id> (opt)Generate standalone HTML/CSS prototype
/dx-figma-verify<id> (opt)Visually verify prototype against Figma reference
Phase 2

Planning

Generate and validate implementation plan.

CommandArgumentDescriptionOutput
/dx-plan<id> (opt)Generate step-by-step implementation plan. Tags steps with repo names for cross-repo work.implement.md
/dx-plan-validate<id> (opt)Verify plan covers all requirementsPass/Fail
/dx-plan-resolve<id> (opt)Fix risks flagged by validationUpdated implement.md
Phase 3

Execution

Execute plan steps, test, review, commit.

CommandArgumentDescription
/dx-step-all<id> (opt)Execute all plan steps autonomously (step + fix loop)
/dx-step<id> (opt)Execute next pending step — implement, test, review, commit
/dx-step-fix<id> (opt)Fix a blocked step — direct fix or corrective steps
Phase 4

Build + Verify

Build, deploy, and run 6-phase verification.

CommandDescription
/dx-step-buildBuild and deploy, auto-fix errors iteratively (max 6 attempts)
/dx-step-verify6-phase verification: compile, lint, test, secrets, architecture, code review
/dx-axe <URL>Accessibility testing using axe MCP Server
AEM

AEM Verification

Component snapshots, verification, and demo capture.

CommandArgumentDescription
/aem-snapshot<component>Baseline component state before changes
/aem-verify<component>Verify component after deployment, create test page
/aem-fe-verify<component>Screenshot and compare against Figma/requirements
/aem-editorial-guide<component>Screenshot dialog, write authoring guide
/aem-qa<id> (opt)Full QA agent — navigate, check rendering, create Bug tickets
/aem-qa-handoff<component> <id>Generate testing steps for QA team, post to ADO
Bug Fix

Bug Fix

Triage, verify, and fix bugs.

CommandArgumentDescription
/dx-bug-all<bug-id>Full bug workflow: triage, verify, fix
/dx-bug-triage<bug-id>Fetch bug, find affected component
/dx-bug-verify<bug-id> [before|after|qa]Reproduce via Chrome DevTools. before: confirm bug. after: verify fix locally. qa: verify fix on QA post-merge.
/dx-bug-fix<bug-id> (opt)Generate fix plan, execute, create PR
Phase 6

Pull Request

Create, review, answer, and fix PRs.

CommandArgumentDescription
/dx-pr<id> (opt)Create pull request via ADO MCP
/dx-pr-commit[pr]Commit with WI linking; add pr to also create a PR
/dx-pr-review<PR URL>Review a PR — analyze diff, post comments, propose fixes. Checks sibling FE repos for field impact on dialog changes.
/dx-pr-review-allnoneBatch-review multiple open PRs
/dx-pr-answer<PR URL> (opt)Answer open PR comments, apply agree-will-fix changes
/dx-pr-review-report<PR URL>Generate categorized report, post to wiki
Phase 7

Documentation

Generate wiki docs, authoring guides, and QA handoffs.

CommandArgumentDescription
/dx-doc-gen<id>Generate wiki docs from spec files, post to ADO wiki
/dx-doc-retro<id>Retroactive wiki docs for completed stories
/aem-doc-gen<id>AEM authoring guide with dialog screenshots
/aem-qa-handoff<component> <id>Generate testing steps for QA team
DoD

Definition of Done

Validate deliverables against the DoD checklist.

CommandArgumentDescription
/dx-req-dod<id>Check Definition of Done and auto-fix gaps
Agents

Agent Roles

Specialized agent pipelines.

CommandArgumentDescription
/dx-agent-all<id> [autonomous]Full pipeline: requirements, plan, execute, review, PR
/dx-agent-re<id>RE Agent — structured requirements spec
/dx-agent-dev<id>Dev Agent — implement from RE spec
Automation

Automation

Deploy and manage autonomous agent infrastructure.

CommandDescription
/auto-initScaffold .ai/automation/
/auto-provisionCreate AWS resources (DynamoDB, SQS, S3, Lambda, API Gateway)
/auto-pipelinesImport ADO pipeline YAMLs, set variables
/auto-deployDeploy Lambda code
/auto-lambda-envSet Lambda environment variables
/auto-webhooksConfigure ADO service hooks + build policy
/auto-alarmsCreate CloudWatch alarms, subscribe email
/auto-doctorFull health check: files, pipelines, Lambda state
/auto-statusDashboard: DLQ depth, token budget, rate limits
/auto-evalRun evaluation framework against test fixtures
/auto-testLocal dry-run against real ADO data
Utility

Utility + Maintenance

Health checks, upgrades, and ejection.

CommandArgumentDescription
/dx-doctor[dx|aem|seed-data|auto|all]Check health of all dx workflow files. Warns on stale project.yaml files.
/dx-upgrade[dx|aem|auto|all]Fix all issues found by dx-doctor. Migrates legacy project.yaml fields into config.yaml.
/dx-eject[dx|aem|auto|all]Eject plugin assets into local repo
/aem-refreshnoneUpdate .ai/project/ seed data
/aem-doctor[components|osgi|dispatcher|all]Check AEM project infrastructure health
Cross-Repo

Cross-Repo Ticket Handling

Some tickets span multiple repos. The workflow handles this gracefully.

How It Works

  1. Requirements phase (/dx-req) discovers which repos need work
  2. research.md and explain.md list the repos involved
  3. Plan and execution only cover the current repo
  4. Switch to the other repo and run the same workflow there

Pipeline Mode (Automated)

When DX_PIPELINE_MODE is set, code-writing coordinators (dx-bug-all, dx-agent-all, dx-req-dod) check if the work targets another repo. If so, they write .ai/run-context/delegate.json and STOP. A post-Claude YAML step reads it and queues the equivalent pipeline in the target repo.

Manual cross-repo workflow
# In backend repo:
/dx-req 2416553
/dx-plan
/dx-step-all

# In frontend repo (same ticket, plans FE steps only):
/dx-req 2416553
/dx-plan
/dx-step-all
Learning

Self-Learning

Coordinator skills automatically log run metrics and patterns. Each subsequent run gets smarter.

FileContentsWritten By
.ai/learning/raw/runs.jsonlRun records — timestamp, ticket, flow, step statuses, fix/heal countsAll three coordinators
.ai/learning/raw/fixes.jsonlFix patterns — error type, fix description, success/failuredx-step-all
.ai/learning/raw/bugs.jsonlBug patterns — component, root cause, files changeddx-bug-all

Pattern Promotion

When dx-step-all detects a fix pattern with 3+ successes and 0 failures, it promotes it to .claude/rules/learned-fix-*.md — auto-loaded by Claude on every future run.

Aggregation Skills

Run /dx-learn to aggregate raw JSONL data into readable markdown summaries. Run /dx-retro for weekly retrospective with success rates, trends, and suggested actions.

Committing Learning Data

.ai/learning/raw/ contains JSONL machine data — safe to commit or gitignore. .claude/rules/learned-*.md should be committed (they are project-level rules).

Workflow

DoR to BA Feedback to Share Plan

Incremental requirements loop with idempotent steps and BA checkbox interaction.

Incremental Requirements Loop

Each step is idempotent -- already-completed steps are skipped on re-run.

req /dx-req Full pipeline
Example workflow
/dx-req 2416553                  # Full pipeline: fetch, DoR, explain, research, share

Dependency Chain

fetchdorexplainresearchshare. The BA checkbox loop between dor steps is the only human gate in the chain. All other transitions are automatic.

KAI by Dragan Filipovic