consult-codex

Dual-AI Consultation: Codex vs Code-Searcher

You orchestrate consultation between OpenAI's Codex and Claude's code-searcher to provide comprehensive analysis with comparison.

When to Use This Skill

High value queries:

• Complex code analysis requiring multiple perspectives
• Debugging difficult issues
• Architecture/design questions
• Code review requests
• Finding specific implementations across a codebase

Lower value (single AI may suffice):

• Simple syntax questions
• Basic file lookups
• Straightforward documentation queries

Workflow

When the user asks a code question:

1. Build Enhanced Prompt

Problem-restate pre-flight (non-blocking). Before building the prompt, emit ONE line restating the code question you are about to dispatch (and, only if genuinely ambiguous, the alternative reading), then proceed:

Reading this as: «one-line restatement» (alt: «other reading», if any) — proceeding to consult; interrupt now to correct the framing.

Emit-and-proceed — do not ask-and-wait (the orchestrator can't reliably detect its own misframing). One line, and it guards the whole dispatch against a wrong-framing run.

Wrap the user's question with structured output requirements:

[USER_QUESTION]

=== Analysis Guidelines ===

**Structure your response with:**
1. **Summary:** 2-3 sentence overview
2. **Key Findings:** bullet points of discoveries
3. **Evidence:** file paths with line numbers (format: `file:line` or `file:start-end`)
4. **Confidence:** High/Medium/Low with reasoning
5. **Limitations:** what couldn't be determined

**Line Number Requirements:**
- ALWAYS include specific line numbers when referencing code
- Use format: `path/to/file.ext:42` or `path/to/file.ext:42-58`
- For multiple references: list each on a SEPARATE line with its own file path
  (avoid comma-separated multi-citation like `file.ts:45, 67, 98`)
- Include brief code snippets for key findings

**Examples of good citations:**
- "The authentication check at `src/auth/validate.ts:127-134`"
- "Configuration loaded from `config/settings.json:15`"
- "Error handling in `lib/errors.ts:45`, `lib/errors.ts:67-72`, and `lib/errors.ts:98`"

**Citations Index (required):** end your response with a fenced block, one line per
Key Finding (repeat each block entry's `file:line` inline in the finding as usual):
```citations
<finding #> — path/to/file.ext:LINE[-END]
```

Severity / no-manufacture block — ORCHESTRATOR-GATED. Append the block below to both agents' prompts identically ONLY when the query is a defect hunt / code review (bug, security audit, "what's wrong with…", "review this"). OMIT it for explanatory / "how does X work" questions, where "found nothing" is not meaningful. The orchestrator — which knows the query type — makes this include/omit decision once, BEFORE writing the prompt files; do not leave it to each agent to self-classify. When included, append exactly these two bullets (the text only — no leading marker):

- Tag each finding with a **Severity** — Critical (wrong/broken on expected inputs) · Warning (fails on unusual but valid inputs) · Info (noteworthy, not actionable). Severity is *impact*, orthogonal to the Confidence field (*certainty*).
- **Finding nothing is a valid, valuable result.** If the code is correct, say so plainly with one verifying note — do NOT manufacture issues to look thorough.

2. Invoke Both Analyses in Parallel

Setup (run first). $CLAUDE_PROJECT_DIR is not always exported into the Bash tool shell, so resolve it with a $PWD fallback and ensure the tmp dir exists. Substitute the resolved literal path for $PROJECT_DIR, and a freshly generated RUN_ID (seconds-resolution + 4-char nonce, e.g. run-2026-05-25-143052-a7f3), into every command below. The RUN_ID in temp filenames prevents collisions between two concurrent invocations sharing $PROJECT_DIR/tmp.

PROJECT_DIR="${CLAUDE_PROJECT_DIR:-$PWD}"
# Validate BEFORE creating tmp — `mkdir -p` would otherwise make the check pass even
# for a bad path (it creates the dir, then `[ -d ]` always succeeds).
[ -d "$PROJECT_DIR" ] || { echo "ERROR: PROJECT_DIR '$PROJECT_DIR' is not a directory" >&2; exit 1; }
mkdir -p "$PROJECT_DIR/tmp"

# Pre-flight (fail fast, not after a 10-min hang). jq is a HARD dependency — output
# parsing needs it — so abort now rather than warn-and-continue into opaque failures.
command -v jq >/dev/null 2>&1 || { echo "ERROR: 'jq' not found — required for output parsing; aborting" >&2; exit 1; }
# codex is a soft dependency — the CODEX_BIN resilience block below resolves or SKIPs it.
command -v codex >/dev/null 2>&1 || \
  zsh  -i -c "type codex" >/dev/null 2>&1 || \
  bash -i -c "type codex" >/dev/null 2>&1 || \
  echo "WARNING: 'codex' not found — will attempt nvm resolution below, else SKIP"

# Sweep stale orphans (>60 min) from crashed prior runs (best-effort, age-based —
# can theoretically delete a live run's files if it paused >60 min; acceptable).
find "$PROJECT_DIR/tmp" -maxdepth 1 -name '*-prompt-*.txt'   -mmin +60 -delete 2>/dev/null
find "$PROJECT_DIR/tmp" -maxdepth 1 -name '*-output-*.jsonl' -mmin +60 -delete 2>/dev/null

Codex binary resilience (run once, before dispatch). An nvm-managed codex can be a symlink whose @openai/codex install is broken (deleted vendor binary → spawn ... ENOENT), and a broken version can sit EARLIER on PATH than a working one. command -v / zsh -i return the broken path, so detect by RUNNING the binary. If the PATH-resolved codex fails, hunt all nvm node installs for one whose --version succeeds and emit its absolute path. Emit CODEX_BIN=SKIP if none work.

CODEX_BIN=""; INTERACTIVE_SHELL=zsh
# Capture WHICH interactive shell resolves codex (nvm may be in only one of
# ~/.zshrc / ~/.bashrc). The dispatch below uses $INTERACTIVE_SHELL so a
# .bashrc-only setup on macOS still works (prior bug: probe accepted bash,
# dispatch hardcoded zsh).
if   zsh  -i -c 'codex --version' >/dev/null 2>&1; then CODEX_BIN="codex"; INTERACTIVE_SHELL=zsh   # codex resolves via zsh
elif bash -i -c 'codex --version' >/dev/null 2>&1; then CODEX_BIN="codex"; INTERACTIVE_SHELL=bash  # codex resolves via bash
else
  # Match symlinks too (-type l): nvm/npm install codex as a bin/ symlink, which -type f misses.
  CODEX_BIN=$(find "$HOME/.nvm/versions/node" -maxdepth 5 -name codex \( -type f -o -type l \) 2>/dev/null | while IFS= read -r p; do
    "$p" --version >/dev/null 2>&1 && { printf '%s\n' "$p"; break; }
  done)
  [ -z "$CODEX_BIN" ] && CODEX_BIN="SKIP"
fi
echo "CODEX_BIN=$CODEX_BIN"                   # MUST echo: shell vars don't persist across Bash tool calls
echo "INTERACTIVE_SHELL=$INTERACTIVE_SHELL"   # the interactive shell that resolves codex; substitute into the dispatch below

# Resolve the timeout binary used to wrap the Codex gen dispatch (§Step 2 below) so a
# hung CLI is bounded rather than running unbounded — the harness may auto-background
# the dispatch, letting it escape the Bash tool's own timeout. Probe BOTH names:
# Homebrew coreutils installs GNU timeout as `gtimeout`; plain `timeout` exists only
# when the coreutils gnubin PATH is on. If neither exists, TIMEOUT_CMD stays empty and
# the dispatch runs UNWRAPPED (best-effort Bash-tool timeout; `brew install coreutils`
# restores the hard guard).
TIMEOUT_CMD=""
if   command -v timeout  >/dev/null 2>&1; then TIMEOUT_CMD="timeout"
elif command -v gtimeout >/dev/null 2>&1; then TIMEOUT_CMD="gtimeout"
fi
echo "TIMEOUT_CMD=$TIMEOUT_CMD"   # substitute into the §Step-2 Codex dispatch (when empty: omit the wrap)

Two-phase dispatch (required). Tool calls in one message run concurrently, so emitting the Codex prompt-file Write and the Codex dispatch together races the dispatch ahead of the file (Codex errors on a missing prompt file). Use two messages: message 1 writes the Codex prompt file (Step 1 below); message 2 issues the Codex dispatch (Step 2) and the Code-Searcher Agent call in parallel:

Gen-dispatch timeout watchdog (GEN_TIMEOUT=1200). Each $TIMEOUT_CMD -k 10 1200-prefixed CLI gen below — SIGTERM at 1200s (20 min), SIGKILL 10s later (-k 10, which also reaps orphaned Node/MCP children) — is bounded against a hung provider CLI that would otherwise run unbounded (the harness may auto-background the dispatch, so the Bash tool's own timeout is not a reliable cap). When TIMEOUT_CMD is empty (no timeout/gtimeout): omit the $TIMEOUT_CMD -k 10 1200 prefix and dispatch unwrapped (the existing §Setup fallback) — set the Bash tool's own timeout parameter to 1300000 ms as a best-effort cap, and brew install coreutils to restore the hard guard. On a timed-out gen (exit 124 = SIGTERM, 137 = SIGKILL): the output file is empty/truncated, so the existing [ -z … ] parse guard already drops the agent — additionally surface Agent X timed out after 1200s (distinct from an auth failure, which leaves non-empty stderr) and re-count against the §Setup minimum-agent guard. Do not retry. Code-Searcher (Agent tool) carries no $TIMEOUT_CMD -k 10 1200 prefix — it is bounded by its own mechanism, not this watchdog.

• For Codex:

  Model ownership — the model is CONFIG-OWNED, never named by this skill. Do not pass -m: inherit the model and reasoning effort from Codex configuration (~/.codex/config.toml — this installation is configured for gpt-5.6-sol, high effort). Report the agent as plain "Codex" everywhere in the report; never a hardcoded version string. A hardcoded label silently misreports the model the moment the config changes: this skill advertised GPT-5.6-terra in seven places while every dispatch had been running gpt-5.6-sol, because the dispatch carries no -m and never did (corrected 2026-08-01).

  Step 1: Write the enhanced prompt to a temp file using the Write tool:

  Write to $PROJECT_DIR/tmp/codex-prompt-RUN_ID.txt with the ENHANCED_PROMPT
  

Content truncated.