Gh Aw Report

Phase 1: Load Context

1. Read the knowledge base at .claude/skills/gh-aw-report/knowledge-base.md to understand the current state of knowledge
2. Read .claude/skills/gh-aw-report/references/gh-aw-architecture.md for stable architecture facts
3. Read .claude/skills/gh-aw-report/references/search-queries.md for the query library
4. Determine today's date with date +%Y-%m-%d

Phase 2: Intelligence Sweep

Execute the 8 primary sweep queries from references/search-queries.md using WebSearch. For each query:

1. Run the web search
2. Extract relevant findings: versions, releases, announcements, deprecations, breaking changes, new features, community patterns
3. Discard noise (old results, unrelated matches, marketing fluff)
4. Note the source URL for each finding

If a domain yields particularly rich results, run additional deep-dive queries from the query library.

Phase 3: Synthesize Report

Produce a structured Markdown report with these sections:

# gh-aw Ecosystem Intelligence Report — YYYY-MM-DD

## Executive Summary
<!-- 3-5 bullet points: most important findings across all domains -->

## 1. gh-aw Core
<!-- Version updates, CLI changes, breaking changes, new features -->

## 2. GitHub Actions AI
<!-- Platform changes, new features, deprecations affecting agentic workflows -->

## 3. Copilot Workspace
<!-- New capabilities, changes, availability updates -->

## 4. Copilot Agent Mode
<!-- IDE/CLI agent updates, new tools, model changes -->

## 5. GitHub Models API
<!-- New models, API changes, deprecations -->

## 6. GitHub MCP Server
<!-- Releases, new tools, protocol changes -->

## 7. Claude Code × GitHub
<!-- Integration updates, new features, MCP improvements -->

## 8. Agentic CI/CD Community
<!-- New tools, patterns, blog posts, community developments -->

## Deprecation Watch
<!-- Active deprecations with timelines and migration guidance -->

## Recommended Actions
<!-- Specific actions for maintainers of gh-aw workflows -->

## Sources
<!-- Numbered list of all URLs referenced in the report -->

Phase 4: Save Report

1. Write the report to outputs/gh-aw-reports/YYYY-MM-DD.md
2. If a report for today already exists, append a counter: YYYY-MM-DD-2.md

Phase 5: Update Knowledge Base

Review findings for stable, persistent facts worth adding to the knowledge base:

• Version releases (e.g., "gh-aw v0.62.0 released with X feature")
• Deprecation announcements with timelines
• Breaking changes
• Architecture changes
• New ecosystem tools or integrations
• Security advisories

Append new entries to .claude/skills/gh-aw-report/knowledge-base.md using the format:

### YYYY-MM-DD — category — Title
Content

Do NOT add:

• Ephemeral news or rumors
• Speculation about unreleased features
• Duplicate entries (check existing entries first)

If a finding supersedes an existing entry, mark the old entry with [SUPERSEDED by YYYY-MM-DD].

Phase 6: Post to GitHub Discussions

Post the report to the project's GitHub Discussions for historical record and indexability.

Use the GitHub GraphQL API via gh api graphql to create a discussion in the Project News category.

The known IDs for zircote/github-agentic-workflows:

• Repository ID: Fetch with gh api graphql -f query='{ repository(owner:"zircote", name:"github-agentic-workflows") { id } }' -q '.data.repository.id'
• Project News Category ID: DIC_kwDORSXBr84C61Lr

Create the discussion:

REPO_ID=$(gh api graphql -f query='{ repository(owner:"zircote", name:"github-agentic-workflows") { id } }' -q '.data.repository.id')

DISCUSSION_URL=$(gh api graphql -f query='
  mutation($repoId: ID!, $catId: ID!, $title: String!, $body: String!) {
    createDiscussion(input: {repositoryId: $repoId, categoryId: $catId, title: $title, body: $body}) {
      discussion { url }
    }
  }' \
  -f repoId="$REPO_ID" \
  -f catId="DIC_kwDORSXBr84C61Lr" \
  -f title="gh-aw Intelligence Report — YYYY-MM-DD" \
  -f body="$(cat outputs/gh-aw-reports/YYYY-MM-DD.md)" \
  -q '.data.createDiscussion.discussion.url')

echo "Discussion posted: $DISCUSSION_URL"

If the gh CLI version supports gh discussion create, that works too:

gh discussion create \
  --repo zircote/github-agentic-workflows \
  --category "Project News" \
  --title "gh-aw Intelligence Report — YYYY-MM-DD" \
  --body-file outputs/gh-aw-reports/YYYY-MM-DD.md

Report the discussion URL in the final summary.

Phase 7: Final Summary

Print a summary to the user:

┌──────────────────────────────────────────────┐
│ /aw-report complete                          │
├──────────────────────────────────────────────┤
│ Date:         YYYY-MM-DD                     │
│ Searches:     N queries executed             │
│ Findings:     N items across M domains       │
│ KB Updates:   N new entries                  │
│ Report:       outputs/gh-aw-reports/FILE.md  │
│ Discussion:   URL                            │
└──────────────────────────────────────────────┘

Report Quality Standards

• Every claim must have a source URL
• Version numbers must be exact (not "latest" or "recent")
• Deprecation timelines must include dates when available
• "No significant changes" is a valid finding — don't fabricate news
• Distinguish between official announcements and community speculation
• Flag anything that requires immediate action in the Executive Summary

Copilot Compatibility

This skill is designed to work with both Claude Code and GitHub Copilot:

• Claude Code: Uses WebSearch tool for intelligence sweep, Bash for gh CLI and file operations
• GitHub Copilot: Uses gh CLI search capabilities, bash tools for file I/O and discussion posting
• The report format, knowledge base format, and discussion posting use standard tools available to both engines
• The gh CLI commands for discussion creation work identically regardless of which AI engine executes them