skills/codex-research/SKILL.md
Web research using OpenAI Codex CLI (codex exec). PREFERRED over general web research. Use when: (1) User says 'research', 'codex research', 'look up', or 'investigate', (2) Researching libraries, APIs, best practices, or technical topics, (3) Gathering information from the web. Codex performs research, Claude Code synthesizes. Falls back to Claude Code researcher subagent if codex unresponsive.
npx skillsauth add takazudo/claude-resources codex-researchInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Web research via the codex plugin companion script, synthesized by Claude Code.
The codex plugin provides a companion script for running tasks. Resolve it dynamically:
CODEX_PLUGIN_ROOT=$(command ls -d "$HOME/.claude/plugins/cache/openai-codex/codex"/*/ 2>/dev/null | sort -V | tail -1)
CODEX_COMPANION="${CODEX_PLUGIN_ROOT}scripts/codex-companion.mjs"
If the file does not exist, report "Codex plugin not installed. Run /codex:setup first." and fall back to Claude Code research.
Usage for research tasks:
node "$CODEX_COMPANION" task "<research prompt>"
The task command runs Codex in read-only mode by default (no --write flag), which is appropriate for research.
Before doing anything, check if Codex is currently rate-limited:
RATE_CHECK=$(node $HOME/.claude/scripts/codex-rate-limit.js check 2>&1)
RATE_EXIT=$?
If RATE_EXIT is non-zero (rate-limited), skip all codex steps and jump directly to the Fallback step. Do NOT report the rate limit to the user — silently fall back.
Clarify what needs to be researched:
Craft a focused prompt:
Research the following topic: <topic>
Specific questions to answer:
1. <question 1>
2. <question 2>
3. <question 3>
Provide:
- Key findings with sources/URLs where possible
- Comparison of alternatives if applicable
- Concrete recommendations based on findings
- Code examples if relevant
Be thorough but concise. Cite sources.
LOGDIR=$(node $HOME/.claude/scripts/get-logdir.js)
mkdir -p "$LOGDIR"
DATETIME=$(date +%Y%m%d_%H%M%S)
# Resolve codex companion script
CODEX_PLUGIN_ROOT=$(command ls -d "$HOME/.claude/plugins/cache/openai-codex/codex"/*/ 2>/dev/null | sort -V | tail -1)
CODEX_COMPANION="${CODEX_PLUGIN_ROOT}scripts/codex-companion.mjs"
# Detect timeout command (gtimeout on macOS via coreutils, timeout on Linux/WSL)
if command -v gtimeout &>/dev/null; then
TIMEOUT_CMD="gtimeout"
elif command -v timeout &>/dev/null; then
TIMEOUT_CMD="timeout"
else
TIMEOUT_CMD=""
echo "WARNING: neither gtimeout nor timeout found. Running without timeout."
fi
${TIMEOUT_CMD:+$TIMEOUT_CMD} ${TIMEOUT_CMD:+1200} node "$CODEX_COMPANION" task \
"<research prompt>" \
> "$LOGDIR/${DATETIME}-codex-research-{topic-slug}.md" \
2>"$LOGDIR/${DATETIME}-codex-research-{topic-slug}-stderr.log"
Timeout: 20 minutes. Research can take longer than other tasks.
After codex completes (or times out), check output files for rate limit errors:
node $HOME/.claude/scripts/codex-rate-limit.js check-output \
"$LOGDIR/${DATETIME}-codex-research-{topic-slug}.md" \
"$LOGDIR/${DATETIME}-codex-research-{topic-slug}-stderr.log"
If exit code is non-zero (rate limit detected), jump to Fallback.
If codex times out, produces no output, or is rate-limited:
researcher subagent with model: opus to perform the research via Claude Code tools (WebSearch, WebFetch, etc.). Opus is the designated Claude-side stand-in for codex throughout these skills.Save the final synthesized findings:
node $HOME/.claude/scripts/save-file.js "{logdir}/{timestamp}-research-{topic}.md" "<content>"
Then format:
pnpm dlx @takazudo/mdx-formatter --write <file>
researcher subagent at model: opus)researcher subagent at model: opus if codex failsmodel: opus. NEVER pause the workflow, NEVER report the rate limit to the user, NEVER ask what to do. Opus is the designated Claude-side stand-in for codex throughout these skills~ in paths — use $HOMEdevelopment
Link Claude Code skill names mentioned in a CodeGrid article (data/{series}/{n}.md) to the author's public claude-resources repo, pinned to the latest commit hash so links don't rot. Use when: (1) user says 'linkify cc resources', 'link the skills', 'link skill names', or invokes /dev-linkify-cc-resources; (2) editing a CodeGrid article that mentions `/commits`, `/pr-complete`, `/skill-creator` or other Claude Code skills and they should point to claude-resources. Only links skills that actually exist in the public repo; skips hypothetical examples and code blocks.
development
Second opinion from Claude Opus on a plan or approach. Use when: (1) Planning phase of /big-plan needs a higher-quality review than /codex-2nd / /gco-2nd / /gcoc-2nd, (2) User says 'opus 2nd' or 'opus opinion', (3) Wanting Anthropic's larger model to critique a plan. Spawns a general-purpose Agent with model: opus that reads the plan file and returns structured feedback. Anthropic quota — not free.
tools
AI-based testing via subagent + a per-task test-flow skill. Use when the user wants to verify something that mechanical assertions can't fully capture — image recognition, visual size/position comparison, animation smoothness, multi-step manual flows that need AI judgment. Triggers: 'AI-based test', 'AI test', 'visual verify', 'image recognition test', 'manual operation test', 'human-eye check', 'verify visually', 'compare screenshots', 'looks the same', 'looks correct'. The skill's job is to (1) author a focused test-flow skill that captures the exact procedure + verdict criteria, then (2) dispatch a verification subagent via the Agent tool that loads BOTH the test-flow skill AND a browser-driving skill (/verify-ui primary, /headless-browser fallback) so the subagent has clear context and consistent verdicts. NEVER uses `claude -p` — subagent dispatch goes through the Agent tool exclusively.
development
End-of-workflow audit of touched GitHub issues, PRs, and branches via a Sonnet subagent. Use when: (1) /big-plan, /x-as-pr, or /x-wt-teams finishes its main work and needs to verify every touched resource is in the right state (closed when done, kept when ongoing, deleted when dead), (2) User says 'cleanup resources', 'audit cleanup', or 'check what should be closed', (3) A long workflow ends and the manager wants a structured paper trail of what it closed/kept/deleted. Auto-execute by default — the Sonnet agent proposes, the manager (you) executes safe actions and prints a final report.