You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A Claude Code user can adopt gh-aw: Claude is a first-class engine, the quick-start includes an ANTHROPIC_API_KEY setup callout, and every documented tool is engine-agnostic (13/14; web-search is the only engine-dependent one). No critical blockers. The dominant gap — persistent across 10 daily runs — is that the exact Claude Code persona (Claude Pro/Max/Teams OAuth users) is silently rejected: CLAUDE_CODE_OAUTH_TOKEN is unsupported and ignored, documented only in auth.mdx:201/faq.md:368, never surfaced in the quick-start. Overall score: 7.5/10 (stable).
Severity Findings
Critical / Major / Minor (with file refs)
🚫 Critical Blockers: None. Claude works end-to-end via engine: claude + ANTHROPIC_API_KEY.
⚠️ Major Obstacles
OAuth token silently ignored — CLAUDE_CODE_OAUTH_TOKEN (Claude Max/Teams/Pro subscription auth, the natural credential for this persona) is not supported and ignored if set. Buried in auth.mdx:201 + faq.md:368; absent from setup/quick-start.mdx. A subscription-only Claude user hits a confusing failure. Fix: add a NOTE in quick-start step ~71 stating only ANTHROPIC_API_KEY or WIF works, not the OAuth token.
gh aw init defaults to Copilot artifacts — generates Copilot dispatcher/agent unless --engine claude is passed (cli.md:133,138), but that flag is never shown in the quick-start steps. Fix: show gh aw init --engine claude in the quick-start.
Omitting engine: silently defaults to Copilot (how-they-work.mdx:26) across 35 example workflows; quick-start inline secret callouts cover only Copilot + Anthropic, not Codex/Gemini (quick-start.mdx:70-71,84-87). Fix: make the default explicit and add per-engine secret callouts.
💡 Minor Confusion
Copilot-first ordering everywhere (README.md:40, how-they-work.mdx:26); no "why Claude vs Copilot" guidance.
custom engine: 0 standalone example workflows (only a shared/genaiscript.md import fragment) and no quick-start auth section.
Codex setup is thin — quick-start.mdx:71 names OPENAI_API_KEY with a link but no step-by-step, unlike Copilot (78-81) and Claude (84-87).
Engine & Tool Matrix
Engine
Setup (quick-start)
Examples (.github/workflows)
Auth
Score
Copilot (default)
✅ default path, org-billing or PAT choice
121
COPILOT_GITHUB_TOKEN or copilot-requests:write, fully documented
Tool classification (reference/tools.md): 13/14 engine-agnostic, 1 unclear (web-search, engine-dependent — Codex disables it by default unless declared, tools.md:62-67). No copilot-only, claude-only, or codex-only tools exist — a strong positive for Claude adoption. Only cross-engine variance is timeout defaults (Claude 60s / Codex 120s, tools.md:157).
Parity: Copilot dominates (~71% of engine-declaring files); Claude has solid coverage (57, near-half of Copilot, concentrated in analysis/security/reporting workflows) plus smoke-test parity (smoke-claude.md). Codex modest (16). Custom has zero standalone runnable parity.
Auth Gaps
Per-engine auth extraction
Engine
Secret
Gap
copilot
COPILOT_GITHUB_TOKEN
distinct from GITHUB_TOKEN; org-billing alternative documented
claude
ANTHROPIC_API_KEY
✅ in quick-start; OAuth token unsupported & not warned in quick-start; no Bedrock/Vertex/model-selection mention
codex
OPENAI_API_KEY
no step-by-step key creation in quick-start
custom
none
no auth/secret/setup documented in quick-start or cli.md
Both setup docs defer full credential specifics to reference/auth.mdx (not in setup/), so a quick-start-only reader misses the OAuth caveat and WIF option entirely.
Recommended Actions
Priority 1 — Add a quick-start NOTE: Claude requires ANTHROPIC_API_KEY (or WIF); CLAUDE_CODE_OAUTH_TOKEN / subscription OAuth is not supported. Closes the single most persona-relevant gap (open 10 runs).
Priority 2 — Surface gh aw init --engine claude in quick-start steps; make the implicit Copilot default explicit; add per-engine (Codex/Gemini) secret callouts.
Priority 3 — Add a "choosing an engine" / why-Claude-vs-Copilot section; add at least one standalone engine: custom example with auth notes; flesh out Codex key-creation steps.
Note: cache-memory (review-history.jsonl) write was blocked by sandbox this run (working-dir-only writes); trend data — score 7.5, 0 critical, identical 3 major gaps — is reported inline above.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
Executive Summary
A Claude Code user can adopt gh-aw: Claude is a first-class engine, the quick-start includes an
ANTHROPIC_API_KEYsetup callout, and every documented tool is engine-agnostic (13/14;web-searchis the only engine-dependent one). No critical blockers. The dominant gap — persistent across 10 daily runs — is that the exact Claude Code persona (Claude Pro/Max/Teams OAuth users) is silently rejected:CLAUDE_CODE_OAUTH_TOKENis unsupported and ignored, documented only inauth.mdx:201/faq.md:368, never surfaced in the quick-start. Overall score: 7.5/10 (stable).Severity Findings
Critical / Major / Minor (with file refs)
🚫 Critical Blockers: None. Claude works end-to-end via
engine: claude+ANTHROPIC_API_KEY.CLAUDE_CODE_OAUTH_TOKEN(Claude Max/Teams/Pro subscription auth, the natural credential for this persona) is not supported and ignored if set. Buried inauth.mdx:201+faq.md:368; absent fromsetup/quick-start.mdx. A subscription-only Claude user hits a confusing failure. Fix: add a NOTE in quick-start step ~71 stating onlyANTHROPIC_API_KEYor WIF works, not the OAuth token.gh aw initdefaults to Copilot artifacts — generates Copilot dispatcher/agent unless--engine claudeis passed (cli.md:133,138), but that flag is never shown in the quick-start steps. Fix: showgh aw init --engine claudein the quick-start.engine:silently defaults to Copilot (how-they-work.mdx:26) across 35 example workflows; quick-start inline secret callouts cover only Copilot + Anthropic, not Codex/Gemini (quick-start.mdx:70-71,84-87). Fix: make the default explicit and add per-engine secret callouts.💡 Minor Confusion
README.md:40,how-they-work.mdx:26); no "why Claude vs Copilot" guidance.customengine: 0 standalone example workflows (only ashared/genaiscript.mdimport fragment) and no quick-start auth section.quick-start.mdx:71namesOPENAI_API_KEYwith a link but no step-by-step, unlike Copilot (78-81) and Claude (84-87).Engine & Tool Matrix
COPILOT_GITHUB_TOKENorcopilot-requests:write, fully documented--engine claudenot surfacedANTHROPIC_API_KEYin quick-start + WIF keyless; ❌ OAuth token unsupportedOPENAI_API_KEY, named not detailedTool classification (
reference/tools.md): 13/14 engine-agnostic, 1 unclear (web-search, engine-dependent — Codex disables it by default unless declared,tools.md:62-67). No copilot-only, claude-only, or codex-only tools exist — a strong positive for Claude adoption. Only cross-engine variance is timeout defaults (Claude 60s / Codex 120s,tools.md:157).Parity: Copilot dominates (~71% of engine-declaring files); Claude has solid coverage (57, near-half of Copilot, concentrated in analysis/security/reporting workflows) plus smoke-test parity (
smoke-claude.md). Codex modest (16). Custom has zero standalone runnable parity.Auth Gaps
Per-engine auth extraction
COPILOT_GITHUB_TOKENGITHUB_TOKEN; org-billing alternative documentedANTHROPIC_API_KEYOPENAI_API_KEYBoth setup docs defer full credential specifics to
reference/auth.mdx(not insetup/), so a quick-start-only reader misses the OAuth caveat and WIF option entirely.Recommended Actions
Priority 1 — Add a quick-start NOTE: Claude requires
ANTHROPIC_API_KEY(or WIF);CLAUDE_CODE_OAUTH_TOKEN/ subscription OAuth is not supported. Closes the single most persona-relevant gap (open 10 runs).Priority 2 — Surface
gh aw init --engine claudein quick-start steps; make the implicit Copilot default explicit; add per-engine (Codex/Gemini) secret callouts.Priority 3 — Add a "choosing an engine" / why-Claude-vs-Copilot section; add at least one standalone
engine: customexample with auth notes; flesh out Codex key-creation steps.Note: cache-memory (
review-history.jsonl) write was blocked by sandbox this run (working-dir-only writes); trend data — score 7.5, 0 critical, identical 3 major gaps — is reported inline above.Beta Was this translation helpful? Give feedback.
All reactions