dx-doc-gen - SKILL.md Agent Skill

name: dx-doc-gen description: Generate wiki documentation from completed spec files — architecture decisions, usage guide, API changes. Posts to ADO Wiki or Confluence depending on provider config. Use as the final step after implementation is done. Invoked automatically by /dx-agent-all Phase 7 and /dx-req-dod. when_to_use: "Use as the final step after implementation to generate wiki documentation. Trigger on 'generate docs', 'create wiki page', 'document this story', 'post to wiki', or when /dx-agent-all reaches Phase 7." argument-hint: "[ADO Work Item ID (optional — uses most recent if omitted)]" allowed-tools: ["read", "edit", "search", "write", "agent", "ado/", "atlassian/"]

You generate wiki-style documentation from completed spec files. The output captures what was built, why, and how to use it. In pipeline/wiki mode, it posts to the correct sprint subfolder in the ADO wiki or Confluence (depending on tracker.provider).

1. Locate the Spec Directory

SPEC_DIR=$(bash .ai/lib/dx-common.sh find-spec-dir <work-item-id-if-provided>)

If the script exits with error, ask the user for the work item ID.

Read .ai/config.yaml for:

tracker.provider — ado (default) or jira
scm.org — ADO org URL
scm.project — ADO project name
scm.wiki-id — ADO wiki identifier (if configured)
scm.wiki-project — project that owns the wiki (may differ from scm.project)
scm.wiki-doc-root — root wiki path for technical documentation
confluence.space-key — Confluence space key (if provider = jira)
confluence.doc-root — parent page title for generated docs (if provider = jira)

2. Read Source Files

Read these files from $SPEC_DIR (all optional — generate from what's available):

raw-story.md — original story for title, ID, acceptance criteria, iteration path
explain.md — distilled developer requirements
share-plan.md — non-technical summary
implement.md — step-by-step plan with status
research.md — codebase context and patterns found
.sprint — sprint name (e.g., Sprint 41) saved by /dx-req (Phase 1)
.branch — feature branch name
.pr — PR number and URL (if PR was created)
demo/authoring-guide.md — AEM authoring guide with screenshots (produced by /aem-doc-gen)
figma-extract.md — Figma design data (produced by /dx-figma-extract)
aem-fe-verify.md — FE verification verdict (produced by /aem-fe-verify)
aem-after.md — post-deployment component state (produced by /aem-verify)
qa-handoff.md — QA testing steps with page URLs (produced by /aem-qa-handoff)

If none exist (excluding authoring-guide.md), print "No spec files found — nothing to generate from" and STOP.

3. Determine Sprint

Read $SPEC_DIR/.sprint for the sprint name. If the file doesn't exist, try to extract it from the Iteration Path in raw-story.md. If still unknown, set sprint to Unknown and warn the user.

4. Check Existing Output

Check if docs/wiki-page.md exists in the spec directory
If it exists, read its content
Check staleness: does the ADO ID match? Has implement.md changed since generation?
If current → print docs/wiki-page.md already up to date — skipping and STOP
If outdated → print docs/wiki-page.md exists but is outdated — regenerating
If not found → continue normally

5. Generate wiki-page.md

Create $SPEC_DIR/docs/ directory if it doesn't exist.

5a. Template — With Authoring Guide (AEM projects)

If demo/authoring-guide.md exists, read .ai/templates/wiki/wiki-page-aem.md.template and follow that structure exactly. The template is structured as a demo walkthrough — sections are ordered to match a typical demo presentation flow (story → design → implementation → QA page → authoring/dialog → frontend → Figma comparison).

Authoring section is the core of the demo page. It merges what was previously split across Configuration, Dialog Changes, and Authoring Guide into one presenter-friendly section. Every action MUST have a clickable link — the AEM Author Edit URL where the component exists, the real component name, and a field-by-field table of what to change. The presenter should be able to follow this section to configure and demo the component live without searching for anything.

Screenshot references — repo-relative paths: ADO wiki does not support binary upload via current MCP tools. All screenshot references MUST use repo-relative paths so they can be committed and linked from the ADO wiki.

Compute <repo-relative-path> = .ai/specs/<id>-<slug> (relative to repo root). Use this for all image references in the generated wiki page:

<repo-relative-path>/demo/dialog-<component>.png
<repo-relative-path>/demo/rendered-<component>.png
<repo-relative-path>/prototype/figma-reference.png
<repo-relative-path>/screenshots/component-rendered.png

Conditional sections: The template contains sections guarded by comments (e.g., "ONLY if figma-extract.md exists"). Check for each source file and OMIT the entire section (heading included) if the source data is not available. Specifically:

Design Reference — only if figma-extract.md or prototype/ dir exists
Figma Comparison — only if BOTH Figma reference AND rendered screenshot exist

QA Verification (MANDATORY for AEM projects): AEM pages are the deliverable — this is what the customer wants to see. Every AEM wiki page MUST include QA page URLs. Read QA URLs from .ai/config.yaml (aem.author-url-qa, aem.publish-url-qa).

Two views are required per page:

Author Edit: <author-url-qa>/editor.html<page>.html — dialog/config view for authors
Author Preview: <author-url-qa><page>.html?wcmmode=disabled — FE rendering without editor chrome, for demo

Page selection rule: New demo pages are ONLY for new components. For updates to existing components (enhancements, a11y fixes), the page should be the best representative existing page that already has the component — not a newly created demo page.

Extract demo page path from (in priority order):

qa-handoff.md — QA-specific URLs
demo/authoring-guide.md — Author URL, Publisher URL
aem-after.md — Test/Demo page path
If none of the above have a page, and this is an existing component update, search for the best representative production page with the component

Multi-component coverage: If several components are affected, determine whether they can all be demoed on one page or need separate pages. Provide a QA URL table for each page needed. If a new component was created but not yet on a QA page, flag: ⚠️ Component <name> needs to be added to a QA page and configured for demo.

5b. Template — Without Authoring Guide (non-AEM or authoring-guide missing)

If demo/authoring-guide.md does NOT exist, read .ai/templates/wiki/wiki-page.md.template and follow that structure exactly. This standard template includes a ## Usage section instead of Authoring/Website.

6. Wiki Mode — Post to Wiki

If the environment variable PIPELINE_MODE is set to true, OR if the user explicitly asks to post to wiki:

Read tracker.provider from .ai/config.yaml (default: ado).

If ado → follow Section 6a (ADO Wiki posting).
If jira → follow Section 6b (Confluence posting).

6a. ADO Wiki Mode

Read scm.wiki-id, scm.wiki-project, and scm.wiki-doc-root from .ai/config.yaml
If scm.wiki-id is not configured, print: Wiki ID not configured — wiki page saved locally only. and STOP wiki posting.
Determine the sprint subfolder path:

WIKI_ROOT = <scm.wiki-doc-root>        # e.g., /My-Wiki/Technical-Documentation/Sprint-wise
SPRINT = <sprint name from .sprint>     # e.g., Sprint 42
PAGE_TITLE = <id>-<slug-from-title>     # e.g., 2357516-Enhance-PLP-Filter-Sticky
WIKI_PATH = "${WIKI_ROOT}/${SPRINT}/${PAGE_TITLE}"

Check if the sprint subfolder exists. Try to get the sprint page:

mcp__ado__wiki_get_page
  wikiIdentifier: <scm.wiki-id>
  project: <scm.wiki-project>
  path: "${WIKI_ROOT}/${SPRINT}"

If the sprint subfolder does NOT exist (404), create it first:

mcp__ado__wiki_create_or_update_page
  wikiIdentifier: <scm.wiki-id>
  project: <scm.wiki-project>
  path: "${WIKI_ROOT}/${SPRINT}"
  content: "# ${SPRINT}\n\nTechnical documentation for stories in ${SPRINT}."

Check for existing page by ID prefix. The work item ID is unique — if a page already exists for this story (possibly with a different slug), reuse its path instead of creating a duplicate:

mcp__ado__wiki_list_pages
  wikiIdentifier: <scm.wiki-id>
  project: <scm.wiki-project>
  path: "${WIKI_ROOT}/${SPRINT}"

Scan the returned child pages for any whose name starts with <id>- (e.g., 2451742-). If found, set WIKI_PATH to that existing page's path — do NOT use the newly generated slug.

Create or update the wiki page:

mcp__ado__wiki_create_or_update_page
  wikiIdentifier: <scm.wiki-id>
  project: <scm.wiki-project>
  path: "${WIKI_PATH}"
  content: <contents of docs/wiki-page.md>

Print: Wiki page created/updated at ${WIKI_PATH}

If the sprint is Unknown, post under ${WIKI_ROOT}/Unsorted/${PAGE_TITLE} instead (apply the same ID-prefix lookup in Unsorted) and warn: Sprint unknown — page placed under Unsorted. Move it manually after confirming the sprint.

6b. Confluence Mode

Read confluence.space-key and confluence.doc-root from .ai/config.yaml.
If confluence.space-key is not configured, print: Confluence space key not configured — wiki page saved locally only. and STOP wiki posting.
Find the doc root parent page:

mcp__atlassian__confluence_search
  cql: "title = '<confluence.doc-root>' AND space = '<confluence.space-key>'"

Extract page_id from results → DOC_ROOT_ID. If not found, print: Doc root page "<confluence.doc-root>" not found in space <space-key> — create it in Confluence first. and STOP.

Find or create the sprint subfolder page:

SPRINT = <sprint name from .sprint>   # e.g., Sprint 42

mcp__atlassian__confluence_search
  cql: "title = '<SPRINT>' AND ancestor = '<DOC_ROOT_ID>' AND space = '<confluence.space-key>'"

If not found, create it:

mcp__atlassian__confluence_create_page
  space_key: "<confluence.space-key>"
  title: "<SPRINT>"
  body: "<h1><SPRINT></h1><p>Technical documentation for stories in <SPRINT>.</p>"
  parent_id: "<DOC_ROOT_ID>"

Save the sprint page ID → SPRINT_PAGE_ID.

Check if the doc page already exists:

PAGE_TITLE = <id>-<slug-from-title>   # e.g., 2357516-Enhance-PLP-Filter-Sticky

mcp__atlassian__confluence_search
  cql: "title = '<PAGE_TITLE>' AND ancestor = '<SPRINT_PAGE_ID>' AND space = '<confluence.space-key>'"

Content format conversion (CRITICAL):

Skills generate markdown (same as for ADO wiki). Before posting to Confluence, test if the MCP server accepts markdown directly by making a test call. If the page renders correctly with raw markdown, use markdown. If it renders as raw text, convert markdown to basic Confluence storage format (XHTML):

# Heading → <h1>Heading</h1>
**bold** → <strong>bold</strong>
*italic* → <em>italic</em>
- item → <ul><li>item</li></ul>
1. item → <ol><li>item</li></ol>
`code` → <code>code</code>
Code blocks → <ac:structured-macro ac:name="code"><ac:plain-text-body><![CDATA[...]]></ac:plain-text-body></ac:structured-macro>
Tables → <table><tr><th>...</th></tr><tr><td>...</td></tr></table>
Links [text](url) → <a href="url">text</a>
Paragraphs → <p>...</p>

7a. If page doesn't exist — create:

mcp__atlassian__confluence_create_page
  space_key: "<confluence.space-key>"
  title: "<PAGE_TITLE>"
  body: "<converted content>"
  parent_id: "<SPRINT_PAGE_ID>"

7b. If page exists — update:

mcp__atlassian__confluence_update_page
  page_id: "<existing page ID>"
  title: "<PAGE_TITLE>"
  body: "<converted content>"
  version_number: <current version + 1>

Print: Wiki page created/updated in Confluence: <confluence.space-key> > <SPRINT> > <PAGE_TITLE>

If the sprint is Unknown, create the page under DOC_ROOT_ID directly with an "Unsorted" parent page (create if needed), and warn: Sprint unknown — page placed under Unsorted. Move it manually after confirming the sprint.

7. Present Summary

## dx-doc-gen complete

**<Title>** (ADO #<id>)
- Sprint: <sprint name>
- Output: `<spec-dir>/docs/wiki-page.md`
- Sections: <list of included sections>
- Template: AEM (demo walkthrough) / Standard (technical doc)
- Screenshots: <count> referenced (<list of screenshot files found>)
- Committable screenshots: `git add <spec-dir>/demo/*.png <spec-dir>/prototype/*.png <spec-dir>/screenshots/*.png`
- Figma: Yes / No
- QA URLs: Yes / No
- Word count: <N> words
- Mode: Local / Wiki (posted to <wiki path>)

Examples

/dx-doc-gen 2416553 — Reads spec files, generates docs/wiki-page.md with Summary, What Changed and Why, Usage, Files Changed. Standard template.
/dx-doc-gen 2416553 (AEM project with authoring guide) — Detects demo/authoring-guide.md. Uses AEM template: Summary → Design Reference → What Changed and Why → QA Verification → Authoring (page URL + component + field table + dialog screenshot) → Frontend Preview → Figma Comparison → Files Changed. Screenshots use repo-relative paths.
/dx-doc-gen 2416553 (pipeline mode) — PIPELINE_MODE=true triggers wiki posting. Creates the sprint subfolder (Sprint 42) in the ADO wiki if it doesn't exist, then creates the wiki page at the configured doc root path. Summary includes git add command for committing screenshots.

Troubleshooting

"No spec files found — nothing to generate from" Cause: The spec directory exists but contains no recognized files (explain.md, implement.md, etc.). Fix: Run the requirements and planning pipeline first (/dx-req <id> or /dx-plan <id>). Even raw-story.md alone is enough for a basic page.
"docs/wiki-page.md already up to date — skipping" Cause: The wiki page was already generated and the source files haven't changed. Fix: Delete the existing docs/wiki-page.md and re-run, or modify one of the source spec files to trigger regeneration.
Wiki posting fails with "Wiki ID not configured" Cause: scm.wiki-id is not set in .ai/config.yaml. Fix: Add scm.wiki-id, scm.wiki-project, and scm.wiki-doc-root to your config. These can be found in ADO under Project Settings > Wiki.

Rules

Read config, never hardcode — ADO URLs, project names, wiki paths from config.yaml
Skip empty sections — never write "N/A" or "None" filler. Omit the section entirely.
Tech lead audience — more detailed than share-plan.md, less raw than explain.md. Include actual techniques and patterns used.
No acceptance criteria — the linked ADO ticket has the full AC. The wiki page links to the ticket, no need to duplicate.
Idempotent — check existing output before regenerating
Degrade gracefully — generate from whatever spec files are available. More files = richer output, but even raw-story.md alone is enough for a basic page.
Sprint from .sprint file — do NOT hardcode sprint numbers. Always read from the file or fall back to iteration path parsing.
Wiki path encoding — ADO wiki uses spaces in paths. Do NOT URL-encode the path when passing to MCP tools.
No time estimates — never include estimates or durations
Authoring-guide drives template — if demo/authoring-guide.md exists, use the AEM demo walkthrough template; otherwise use the standard template with Usage section
Repo-relative screenshot paths — all image references use .ai/specs/<id>-<slug>/... paths so screenshots can be committed to the repo and linked from ADO wiki
Omit empty conditional sections — Design Reference, Figma Comparison are conditional. If source data is missing, omit the entire section including heading. Never write "N/A" placeholders.
QA pages are mandatory — for AEM projects, QA Author Edit + Preview URLs are required. This is the deliverable. Preview uses wcmmode=disabled for FE demo.
Multi-component = multi-page — if several components are affected and need different pages, provide QA URL tables for each page
Don't list every file — group by area (FE/BE/Config) with short descriptions. The PR has the diff; the wiki explains what and why.
Demo walkthrough order — AEM template sections follow a demo presentation flow: story → design → implementation → QA → authoring (page URL + component + fields + dialog screenshot) → frontend → Figma comparison → files
Authoring = presenter runbook — every action in the Authoring section MUST have a clickable AEM Author Edit URL and the real component name. The presenter should never have to search for anything — all links, component names, and field values are explicit.