[claude-code-user-docs-review] 🔍 Claude Code User Documentation Review - 2026-07-11 #44952
Closed
Replies: 1 comment
-
|
This discussion was automatically closed because it expired on 2026-07-12T12:55:12.125Z.
|
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Executive Summary
A Claude Code user who avoids Copilot can onboard fully on standard GitHub with zero Copilot dependency: pass
--engine claudeand setANTHROPIC_API_KEY(or configure WIF keyless). No tool is Copilot-, Claude-, or Codex-exclusive. Overall score 7.5/10, 0 critical blockers — stable across 25 consecutive runs. The single most persona-relevant friction remains:CLAUDE_CODE_OAUTH_TOKENis explicitly unsupported/silently-ignored and this caveat never appears in the quick-start — a Claude Code user's most likely wrong assumption surfaces only after failure.Severity Findings
Critical / Major / Minor (click to expand)
Critical Blockers: none. Every Copilot step has a working Claude-equivalent path; capability parity is intact.
Major Obstacles
CLAUDE_CODE_OAUTH_TOKEN(fromclaude setup-token, Claude Max/Teams billing) is not supported and silently ignored —docs/reference/auth.mdx:206-208, restated at troubleshootingauth.mdx:356. OnlyANTHROPIC_API_KEY(or Anthropic WIF) works. The quick-start Claude tab (quick-start.mdx:97-101) shows only the static console key, so the exact persona hits this only after a run fails. Copilot, by contrast, has a genuine PAT-free path viacopilot-requests: write+${{ github.token }}(auth.mdx:15).gh aw initdefaults to Copilot artifacts; engine omission silently defaults to Copilot.gh aw initemits Copilot-specific artifacts unless--engine claude(cli.md:142); an omittedengine:field makes the runtime default to Copilot (37 of 256 workflow files). This consequence is disclosed only atquick-start.mdx:75, not in Prerequisites — a Claude-only user who removes/omitsengine:gets a Copilot run that fails on their Anthropic key.quick-start.mdx:30vs:97-112). WIF keyless auth (auth.mdx:214-278) — the closest thing to a keyless Claude path — is never surfaced in the quick-start or the add-wizard.Minor Confusion
quick-start.mdx:30). No "why Claude vs Copilot" guidance.shared/genaiscript.md); no default Claude model/version documented (architecture.mdx:280);crushappears in the CLI secrets-bootstrap list (cli.md:260) but is otherwise undocumented.Engine & Tool Matrix
copilot-requests:writeorCOPILOT_GITHUB_TOKENsmoke-claude.md)ANTHROPIC_API_KEYor WIF; OAuth token ignoredOPENAI_API_KEY/CODEX_API_KEYshared/genaiscript.mdonly)engine.envbase-url, no quick-start sectionParity observations: Copilot outnumbers Claude ~2.6:1 including silent defaults, but Claude leads the inline-string
engine: claudeform (45 vs 29) and has a dedicated smoke test — Claude is first-class, not an afterthought. Tools: 11/12 engine-agnostic, 0 engine-exclusive. Onlyweb-searchis engine-dependent (Codex disabled-by-default, others via MCP) —tools.md:56-68. No Copilot-only, Claude-only, or Codex-only tool exists — a strong parity positive.Auth Gaps
CLAUDE_CODE_OAUTH_TOKENunsupported/ignoredauth.mdx:206-208,:356auth.mdx:214-278quick-start.mdx:30Recommended Actions
Priority 1 — Add a one-line note to the quick-start Claude auth tab (
quick-start.mdx:97-101): "CLAUDE_CODE_OAUTH_TOKENand Claude Max/Teams OAuth billing are not supported — useANTHROPIC_API_KEY." This closes the persona's #1 friction point at the exact moment of setup.Priority 2 — In Prerequisites/Step 2, state that omitting
engine:(or runninggh aw initwithout--engine) defaults to Copilot and will fail with only an Anthropic key. Surface WIF keyless as the Claude "no-static-secret" option with a link toauth.mdx:214-278.Priority 3 — Add a Gemini quick-start auth block for parity; add at least one custom-engine standalone example; document the default Claude model and the
crushengine.References: §29152985138 · §29094634866 · §29021096731
Trend: stable vs 2026-07-10 — score 7.5, 0 critical (25 runs running). Identical 3 major gaps persist; auth.mdx re-read this run confirmed the OAuth-token caveat verbatim.
Beta Was this translation helpful? Give feedback.
All reactions