[claude-code-user-docs-review] 🔍 Claude Code User Documentation Review - 2026-05-19

[github-actions[bot]]

Executive Summary

A developer who uses Claude Code (and not GitHub Copilot) can successfully adopt gh-aw today — the project is fundamentally engine-agnostic, Claude is a first-class engine in the schema, and engine: claude workflows are well represented in the repo (61 example workflows). However, the onboarding path is Copilot-first: the Quick Start defaults to Copilot, the wizard's first option is Copilot, gh aw init scaffolds Copilot-specific files, and a few power features (custom agents, harness scripts, autopilot continuations, web-authoring via Copilot Chat) are intentionally Copilot-only without a documented Claude equivalent.

Key Finding: ✅ Adoption is viable but requires deliberate detour — a Claude Code user has to ignore Copilot-shaped defaults at three or four moments in onboarding. The single hardest blocker is that CLAUDE_CODE_OAUTH_TOKEN is explicitly unsupported, so users who only have a Claude Max / Teams subscription (no pay-per-token API access) cannot use gh-aw without separately provisioning an Anthropic API key.

---

Persona Context

I reviewed this documentation as a developer who:

• ✅ Uses GitHub for version control
• ✅ Uses Claude Code as primary AI assistant
• ❌ Does NOT use GitHub Copilot
• ❌ Does NOT use Copilot CLI
• ❌ Does NOT have a Copilot subscription

---

Question 1: Onboarding Experience

Can a Claude Code user understand and get started with gh-aw?

Yes, but with deliberate detours. The Quick Start guide (docs/src/content/docs/setup/quick-start.mdx) lists Anthropic Claude as a valid AI account in Prerequisites (line 29) and the add-wizard Step 2 explicitly says "Select an AI Engine — Choose between Copilot, Claude, Codex, or Gemini" (line 70). That's good. But every subsequent default leans Copilot.

Specific Issues Found:

• Quick Start asymmetry (docs/src/content/docs/setup/quick-start.mdx lines 75–86): The COPILOT_GITHUB_TOKEN note is 3 detailed steps with a deep-link to a pre-filled PAT creation page; the ANTHROPIC_API_KEY note is 2 steps. A Claude-first user reads less detail than a Copilot user. Minor — but signals priorities.
• Step 4 example only sets engine after the fact (line 130–136): The customization step demos engine: claude in the frontmatter as an optional edit after you've already gone through the wizard, instead of showing how to make Claude the default up front.
• add-wizard interactivity is opaque in docs: There's no documented --engine claude flag mention in the Quick Start (although add-wizard does accept --engine/-e per cli.md line 158). A user reading only the Quick Start has no way to skip the Copilot-default ordering.
• "What's next?" links push back to Copilot examples (line 150): "Peli's Agent Factory" is the named next stop. Nothing flags which examples there are Claude vs. Copilot.
• The sample workflow itself defaults to Copilot: githubnext/agentics/daily-repo-status (the workflow installed in Step 2) does not pin engine:, so it inherits the Copilot default. A user choosing Claude in the wizard must trust the wizard rewrites this — there's no visible confirmation in docs.

Recommended Fixes:

• Add a "Using Claude?" callout at the top of quick-start.mdx linking to a one-paragraph Claude variant.
• Match the ANTHROPIC_API_KEY note depth to the Copilot note (mention the Anthropic Console quickstart link, free-tier notes, what subscription levels work, billing implications).
• In Step 4, lead with engine: claude as a first-class setup choice, not as an after-the-fact customization.
• Document the add-wizard --engine claude shortcut directly in the Quick Start so Claude users can skip the Copilot-shaped wizard flow.

---

Question 2: Inaccessible Features for Non-Copilot Users

What features or steps don't work without Copilot?

A small but real set of features is Copilot-exclusive. The docs are mostly honest about this in the engine feature comparison table (engines.md lines 34–45), but the implications aren't always called out in the surrounding prose.

Features That Require Copilot:

• engine.agent — custom agent files (engines.md line 41, 111–121): The .github/agents/<name>.agent.md pattern is Copilot-only. Claude users cannot reference custom agent personas this way. This is the biggest non-obvious gap.
• engine.harness — custom harness scripts (engines.md lines 293–315): Copilot-only by current design, stated explicitly.
• max-continuations — autopilot mode (engines.md line 38): Copilot-only. Multi-run autopilot does not exist for Claude. Claude has max-turns instead — different semantics.
• Web authoring via /agent agentic-workflows in Copilot Chat (creating-workflows.mdx lines 19–24, 102–120): The "GitHub Web Interface" section explicitly requires Copilot. gh aw init is described as creating a "Copilot coding agent" dispatcher (line 104, line 118). No Claude equivalent is documented.
• gh aw init Copilot bias (cli.md line 132–145, creating-workflows.mdx lines 102–120): This command "creates the dispatcher agent file (.github/agents/agentic-workflows.agent.md)" — which is a Copilot custom agent. For a Claude user, the file is dead weight; the docs do not explain whether to skip init, run it anyway, or use a flag.
• BYOK Mode via Copilot CLI (engines.md lines 201–268): Routing requests to Anthropic through COPILOT_PROVIDER_* vars exists as an alternative path, but it's framed as Copilot-engine routing through external providers. A Claude Code user wouldn't naturally start here.

Features That Work Without Copilot (Engine-Agnostic):

• ✅ All gh aw CLI subcommands (compile, run, status, logs, audit, trial, new, deploy, update, etc.)
• ✅ engine: claude workflows with tools: allowlist enforcement (in acceptEdits mode)
• ✅ All tools: — bash, edit, github, web-fetch, web-search, playwright, cache-memory, repo-memory, qmd, agentic-workflows
• ✅ Custom MCP servers (mcp-servers:)
• ✅ Safe outputs (issues, PRs, comments, labels, discussions)
• ✅ All security features (firewall, sandbox, threat detection, integrity filtering, secret redaction)
• ✅ GitHub App authentication for tool/safe-output writes (only COPILOT_GITHUB_TOKEN must be a PAT)
• ✅ --engine claude override on compile, new, run, trial, validate, add, add-wizard

Missing Documentation for Claude Users:

• No "set Claude as default" guide. Copilot is described as "default" everywhere (engines.md line 24, quick-start.mdx line 70, table line 16). There's no documented way to make Claude the repo-wide default — users have to add engine: claude to every workflow individually.
• No "should I run gh aw init if I'm not using Copilot?" guidance. The command creates Copilot-specific files. Docs imply Claude users should still run it for the MCP integration, but never say so explicitly.
• No CLAUDE.md integration guide. Claude's bare: true toggle (engines.md line 332) suppresses CLAUDE.md memory loading, but the docs don't explain how Claude users should structure their CLAUDE.md for use with gh-aw, or which files conflict with .github/AGENTS.md.
• Claude tool-permission semantics are buried (engines.md lines 439–472): The fact that granting unrestricted bash silently disables --allowed-tools enforcement in Claude is a real security gotcha — it should appear in the Quick Start security callout, not just deep in the engine reference.

---

Question 3: Documentation Gaps and Assumptions

Where does the documentation assume Copilot usage?

The codebase is intentionally engine-agnostic in implementation but Copilot-first in narrative. Most assumptions are linguistic/ordering rather than functional.

Copilot-Centric Language Found In:

• README.md line 40: "Supports GitHub Copilot, Claude (Anthropic), Codex (OpenAI), and Gemini (Google) — pick whichever AI account you already have." Copilot is named first. Acceptable, but consistent throughout.
• docs/src/content/docs/setup/quick-start.mdx lines 75–86: Copilot setup note is more detailed than the Claude note (3 vs 2 steps; deep-link to pre-filled PAT creation page only for Copilot).
• docs/src/content/docs/setup/creating-workflows.mdx lines 19–24: "If you have access to GitHub Copilot, you can create and edit Agentic Workflows directly from the Web Interface." — sole web-authoring path, no Claude/VS Code equivalent shown.
• docs/src/content/docs/setup/creating-workflows.mdx line 96: "If not using Copilot, also adjust the engine: field" — Claude is the exception case, not the default.
• docs/src/content/docs/setup/creating-workflows.mdx lines 102–120: gh aw init section is framed entirely around "Copilot coding agent" enablement.
• docs/src/content/docs/introduction/how-they-work.mdx line 26: "Workflows support GitHub Copilot (default), Claude by Anthropic, Codex, and Gemini by Google." Copilot bolded with "(default)" tag.
• docs/src/content/docs/reference/engines.md line 28: "Copilot supports the broadest gh-aw feature set, including custom agents and autopilot-style continuations". True statement, but it primes readers to default to Copilot for max capability rather than pick by fit.
• Experimental engines using COPILOT_GITHUB_TOKEN (engines.md lines 20–22): Crush, OpenCode, and Pi — three non-Copilot engines — all use COPILOT_GITHUB_TOKEN as their secret. Confusing for a non-Copilot user: "Why does the OpenCode engine need a Copilot token?" Brief explanation would help.

Missing Alternative Instructions:

• No Claude-equivalent of the "GitHub Web Interface" workflow creation section in creating-workflows.mdx. The web-authoring story is Copilot-only by design, but the docs should explicitly say "Use VS Code with Claude Code instead" with linked steps.
• No analog to the agentic-workflows.agent.md dispatcher for Claude/Claude Code. Claude Code reads CLAUDE.md, AGENTS.md, and skill files — there's no documentation showing how to wire those up to mirror the Copilot dispatcher experience.
• No "Why Claude?" subsection in engines.md. The "Which engine should I choose?" paragraph (line 28) gives one-sentence positioning per engine, which is thin.
• CLAUDE_CODE_OAUTH_TOKEN rejection is buried (auth.mdx lines 121–123). This is the single most likely surprise for a Claude subscriber — it deserves a top-of-page callout, not a paragraph 100+ lines down.

---

Severity-Categorized Findings

🚫 Critical Blockers (Score: 1/10 — mostly clear)

Blocker 1: Claude Max / Teams subscription users cannot authenticate

Impact: A subset of Claude Code users (those on Claude Max / Teams subscriptions, no pay-per-token API key) cannot use gh-aw at all without separately obtaining an Anthropic API key.

Current State: docs/src/content/docs/reference/auth.mdx lines 121–123 state: "CLAUDE_CODE_OAUTH_TOKEN is not supported... Provider-based OAuth authentication (such as billing through a Claude Teams or Claude Max subscription) is not supported."

Why It's a Blocker: Many heavy Claude Code users only have subscription access, not Console API key access. They will assume their Claude Code auth carries over and silently fail.

Fix Required: Surface this constraint at the top of auth.mdx, in the Quick Start prereqs, and in the engines reference table. Add a sentence in the Quick Start ANTHROPIC_API_KEY callout: "Claude Max / Teams subscriptions do not provide API key access — you must provision a separate pay-as-you-go API key via the Anthropic Console."

Affected Files: docs/src/content/docs/reference/auth.mdx, docs/src/content/docs/setup/quick-start.mdx, docs/src/content/docs/reference/engines.md

⚠️ Major Obstacles (Score: 4/10 — moderate friction)

Obstacle 1: No documented path to make Claude the repo default

Impact: Every workflow file requires a manual engine: claude line, and the Copilot default is hard-coded into the wizard prompts and example workflows pulled from githubnext/agentics.

Current State: Engines reference (engines.md line 24) says: "Copilot CLI is the default — engine: can be omitted when using Copilot." There is no documented way to flip the default at the repo level.

Suggested Fix: Document a .github/aw/config.yml (or equivalent) default-engine setting, OR explicitly state that the canonical way is to set engine: claude in every workflow and provide a gh aw fix codemod to bulk-rewrite.

Affected Files: docs/src/content/docs/reference/engines.md, docs/src/content/docs/setup/quick-start.mdx

Obstacle 2: `gh aw init` Copilot bias is undocumented

Impact: Claude users running gh aw init get a .github/agents/agentic-workflows.agent.md Copilot dispatcher file with no explanation of whether it's needed or harmful.

Current State: cli.md line 134 says init "creates the dispatcher agent file (.github/agents/agentic-workflows.agent.md)". creating-workflows.mdx line 118 elaborates: "registers the agentic-workflows custom agent in GitHub Copilot". No guidance for Claude users.

Suggested Fix: Add an explicit note: "If you're not using Copilot, this file is harmless but unused. Consider gh aw init --no-mcp or skip the command entirely. For Claude Code integration, see [link to a new Claude-integration guide]."

Affected Files: docs/src/content/docs/setup/cli.md, docs/src/content/docs/setup/creating-workflows.mdx

Obstacle 3: Web-authoring path is Copilot-only with no alternative shown

Impact: Claude users see "If you have access to GitHub Copilot, you can create and edit Agentic Workflows directly from the Web Interface" and have no parallel section. The VS Code/CLI section is engine-agnostic but presented as a fallback, not as the recommended path.

Current State: creating-workflows.mdx lines 19–69 are Copilot-only. Lines 70+ are agent-agnostic but assume a coding agent in the terminal.

Suggested Fix: Reorder sections to lead with "Using a Coding Agent (Claude Code, Codex, Copilot CLI)" since that's engine-neutral, then make the GitHub Web Interface section explicitly Copilot-exclusive in its header.

Affected Files: docs/src/content/docs/setup/creating-workflows.mdx

Obstacle 4: Claude's bypassPermissions security gotcha is buried

Impact: A Claude user who grants bash: "*" for convenience silently loses tool allowlisting — --allowed-tools is ignored in bypassPermissions mode. This is a real security consequence not flagged in the Quick Start or main Tools reference.

Current State: Documented at engines.md lines 439–472, but only in the engine-specific deep dive. Tools reference doesn't mention it.

Suggested Fix: Add a callout in docs/src/content/docs/reference/tools.md and a security note in the Quick Start security section.

Affected Files: docs/src/content/docs/reference/tools.md, docs/src/content/docs/introduction/architecture.mdx

💡 Minor Confusion Points (Score: 5/10 — paper cuts)

• Issue 1: CLAUDE_CODE_OAUTH_TOKEN non-support buried in middle of auth doc. File: docs/src/content/docs/reference/auth.mdx line 121.
• Issue 2: Experimental engines (Crush, OpenCode, Pi) using COPILOT_GITHUB_TOKEN is surprising; no explainer. File: docs/src/content/docs/reference/engines.md lines 20–22.
• Issue 3: "Setting up COPILOT_GITHUB_TOKEN?" callout in Quick Start is more detailed than the Claude callout. File: docs/src/content/docs/setup/quick-start.mdx lines 75–86.
• Issue 4: README line 40 mentions "pick whichever AI account you already have" but Quick Start prerequisites still order Copilot first. File: README.md line 40 vs quick-start.mdx line 29.
• Issue 5: "Which engine should I choose?" gives one-line per engine; Claude positioning ("stronger control over turn limits") is technically correct but reads like a niche feature rather than a primary differentiator. File: docs/src/content/docs/reference/engines.md line 28.
• Issue 6: Sample workflow githubnext/agentics/daily-repo-status doesn't pin engine:, so Claude users must trust the wizard to inject it. File: docs/src/content/docs/setup/quick-start.mdx line 62.
• Issue 7: No explicit explanation that the copilot tool name in cli-proxy examples (tools.md line 138) refers to safeoutputs/mcpscripts CLIs, not the Copilot engine.

---

Engine Comparison Analysis

Available Engines

Based on docs/src/content/docs/reference/engines.md lines 14–22:

• engine: copilot — Fully documented, default, dedicated Quick Start path
• engine: claude — Fully documented, dedicated auth section, first-class feature support (max-turns)
• engine: codex — Fully documented, dedicated auth section, web-search opt-in
• engine: gemini — Fully documented, dedicated auth section
• engine: crush (experimental) — Documented but uses COPILOT_GITHUB_TOKEN confusingly
• engine: opencode (experimental) — Documented but uses COPILOT_GITHUB_TOKEN
• engine: pi (experimental) — Documented; provider-aware secret behavior

No engine: custom exists — instead, BYOK / custom endpoint configuration is documented under engine.api-target and engine.env overrides.

Documentation Quality by Engine

Engine	Setup Docs	Examples	Auth Docs	Overall Score
Copilot	⭐⭐⭐⭐⭐	⭐⭐⭐⭐⭐ (126 in repo)	⭐⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Claude	⭐⭐⭐⭐	⭐⭐⭐⭐ (61 in repo)	⭐⭐⭐⭐	⭐⭐⭐⭐
Codex	⭐⭐⭐⭐	⭐⭐⭐ (13 in repo)	⭐⭐⭐⭐	⭐⭐⭐⭐
Gemini	⭐⭐⭐	⭐ (1 in repo)	⭐⭐⭐⭐	⭐⭐⭐
Crush/OpenCode/Pi	⭐⭐ (experimental)	⭐	⭐⭐ (Copilot-token reliance)	⭐⭐

---

Tool Availability Analysis

Engine-Agnostic Tools (work with any engine):

• edit, bash, github, web-fetch, playwright, cache-memory, repo-memory, qmd, agentic-workflows, cli-proxy, mcp-servers (custom)

Engine-Specific Tool Behaviors:

• web-search — built-in for Codex (opt-in); requires third-party MCP server on other engines (Claude included). See tools.md lines 66–67 and engines.md line 40.
• engine.agent (custom agent file) — Copilot only.
• engine.harness — Copilot only.

Unclear/Undocumented:

• Whether agentic-workflows MCP tool integration works equally well with Claude Code's MCP support
• Whether the cli-proxy tool list has any engine-specific behavior

---

Authentication Requirements

Current Documentation Coverage

Quick Start auth coverage for:

• ✅ Copilot — detailed instructions, deep-link to pre-filled PAT page
• 🟡 Claude — present but thinner (2 steps vs 3)
• ❓ Codex — covered in auth.mdx but not in Quick Start callouts
• ❓ Gemini — covered in auth.mdx but not in Quick Start callouts

Missing for Claude Users

• Top-level callout that CLAUDE_CODE_OAUTH_TOKEN is rejected
• Explicit note that Claude Max/Teams subscriptions cannot be used for billing — must provision a pay-per-token API key
• Pricing/billing pointer for Anthropic Console — Copilot has the PAT permission link; Claude users have no equivalent quick-action link
• Free-tier vs paid-tier guidance for getting started

Secret Names — Verified from Docs

• Copilot: COPILOT_GITHUB_TOKEN (PAT, Copilot Requests:Read) — documented in auth.mdx lines 76–99
• Claude: ANTHROPIC_API_KEY — documented in auth.mdx lines 103–125
• Codex: OPENAI_API_KEY (or CODEX_API_KEY alt) — documented in auth.mdx lines 129–167
• Gemini: GEMINI_API_KEY — documented in auth.mdx lines 171–186

---

Example Workflow Analysis

Workflow Count by Engine (from .github/workflows/*.md in this repo)

copilot:   126 workflows
claude:     61 workflows
codex:      13 workflows
gemini:      1 workflow
crush:       1 workflow
opencode:    1 workflow
pi:          1 workflow
(no engine field, defaults to copilot): 25 workflows

Quality of Examples

Copilot Examples: Plentiful and varied — covers triage, reporting, scanning, performance analysis, CI doctoring, etc. New users get a strong sense of what's possible.

Claude Examples: Strong coverage — 61 workflows including this very review workflow (claude-code-user-docs-review.md), daily-doc-healer, daily-team-evolution-insights, ci-doctor, etc. Sufficient for adoption, but the Quick Start does not point to any of them as a "first Claude workflow to try." A Claude user reading the docs has to discover them on their own.

Codex/Gemini Examples: Codex has 13 examples covering specific use cases. Gemini has a single example — under-represented.

---

Recommended Actions

Priority 1: Critical Documentation Fixes

1. Add CLAUDE_CODE_OAUTH_TOKEN warning at top of auth.mdx — Surface the subscription-vs-API-key constraint before users hit it. File: docs/src/content/docs/reference/auth.mdx
2. Add a Claude-equivalent setup callout to Quick Start — Match the depth of the Copilot callout. File: docs/src/content/docs/setup/quick-start.mdx lines 75–86
3. Document gh aw init behavior for non-Copilot users — Explain whether to run it, skip it, or use --no-mcp. File: docs/src/content/docs/setup/cli.md line 132, docs/src/content/docs/setup/creating-workflows.mdx lines 102–120

Priority 2: Major Improvements

1. Add a "Using Claude Code? Start here" sidebar/callout in quick-start.mdx linking to a Claude-first variant of the guide.
2. Reorder creating-workflows.mdx to lead with engine-neutral coding-agent path and flag the GitHub Web Interface section as Copilot-exclusive.
3. Add a Claude security note about bypassPermissions mode and bash: "*" interaction to tools.md and the Quick Start.
4. Document a "set Claude as default for this repo" pattern — even if it's "add engine: claude to every workflow," say so explicitly.
5. Add a brief explainer for why experimental engines (Crush/OpenCode/Pi) use COPILOT_GITHUB_TOKEN in the engines table.

Priority 3: Nice-to-Have Enhancements

1. Add a "Claude Code integration guide" — show how to wire up CLAUDE.md, .github/AGENTS.md, and bare: true for Claude Code users who already have project memory.
2. Add a "Why Claude?" subsection in engines.md expanding the one-sentence positioning into 2–3 concrete strengths.
3. Add more Gemini examples to bring it in line with Codex.
4. Tag example workflows by engine in the Agent Factory blog so Claude users can filter to Claude examples.
5. Add a gh aw new --engine claude first-class example in the Quick Start so the Claude path doesn't feel grafted on.

---

Positive Findings

What Works Well

• ✅ Engine-agnostic CLI: Every gh aw command supports --engine claude consistently — solid orthogonal design.
• ✅ Claude is a first-class engine in the schema: engine: claude is not a second-class citizen — it appears in every feature comparison table.
• ✅ max-turns documented as Claude-unique strength: The docs are explicit about Claude having a feature Copilot doesn't (engines.md line 37).
• ✅ Dedicated ANTHROPIC_API_KEY auth section: Setup, custom endpoint config, and troubleshooting all covered (auth.mdx lines 103–125).
• ✅ 61 Claude example workflows in the repo: Plenty of patterns to copy.
• ✅ Engine-neutral security architecture: All security features (firewall, threat detection, integrity filtering, safe outputs) work identically across engines — the security story is not Copilot-coupled.
• ✅ GitHub App authentication path is engine-neutral: Only COPILOT_GITHUB_TOKEN is locked to PAT; everything else can use a GitHub App.
• ✅ gh aw secrets bootstrap --engine claude exists and is documented (cli.md line 227) — a Claude user can interactively provision exactly what they need.
• ✅ Honest about Copilot-only features: The engine feature comparison table doesn't pretend Claude has engine.agent or max-continuations.
• ✅ Custom endpoint support for Claude (ANTHROPIC_BASE_URL) — Claude users behind a corporate proxy or using an Anthropic-compatible internal router have a documented path.

---

Conclusion

Can Claude Code Users Successfully Adopt gh-aw?

Answer: ✅ Yes, with deliberate detours. Adoption is viable today; the friction is real but bounded.

Reasoning: Functionally, gh-aw is engine-neutral — Claude is a first-class engine with strong feature support, plenty of examples, dedicated auth docs, and engine-agnostic security. The friction is narrative: every starting-point document (README, Quick Start, Creating Workflows, How They Work) leads with Copilot defaults. A Claude Code user has to consciously override these defaults at 3–4 moments during onboarding (engine selection in wizard, frontmatter edits, gh aw init interpretation, sample-workflow inheritance). None of these are showstoppers, but cumulatively they signal "Copilot is the recommended path," which is misleading given Claude's genuinely first-class support.

The one true blocker is hidden: Claude Max / Teams subscribers without separate Anthropic API access will silently fail to authenticate. This is documented but not surfaced where they'll see it.

Overall Assessment Score: 7.5/10

Breakdown:

• Clarity for non-Copilot users: 6/10 (engine-neutral CLI but Copilot-shaped narrative)
• Claude engine documentation: 8/10 (dedicated sections, but thinner than Copilot)
• Alternative approaches provided: 7/10 (covered, but often as "if not using Copilot, also do X" exception cases)
• Engine parity: 9/10 (real parity in features; small gaps honestly documented)

Next Steps

The cheapest, highest-impact fixes are documentation reordering and callouts — not new content. A pass that:

1. Surfaces the CLAUDE_CODE_OAUTH_TOKEN warning at the top of auth.mdx,
2. Adds a "Using Claude?" sidebar to the Quick Start, and
3. Explains gh aw init for non-Copilot users

...would close ~80% of the gap with a single small PR. Beyond that, a dedicated "Adopting gh-aw with Claude Code" guide would make this project genuinely engine-neutral in tone, not just in implementation.

---

Appendix: Files Reviewed

Complete List of Documentation Files Analyzed

• README.md
• docs/src/content/docs/setup/quick-start.mdx
• docs/src/content/docs/setup/cli.md
• docs/src/content/docs/setup/creating-workflows.mdx
• docs/src/content/docs/introduction/how-they-work.mdx
• docs/src/content/docs/introduction/architecture.mdx
• docs/src/content/docs/reference/tools.md
• docs/src/content/docs/reference/engines.md
• docs/src/content/docs/reference/auth.mdx

Plus repo enumeration: 231 workflow files under .github/workflows/*.md classified by engine: field.

---

Report Generated: Workflow run §26100374617
Workflow: claude-code-user-docs-review
Engine Used: claude (eating our own dog food 🐕)

Generated by 📝 Claude Code User Documentation Review · ● 11.4M · ◷

• expires on May 20, 2026, 1:37 PM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Claude Code User Documentation Review.

A newer discussion is available at Discussion #33546.