Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Consult Codex as an independent expert. Sends a question or task to codex exec and returns the response.
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | — | — |
| case-13 | ✗→✓ | ▲ Improved | — | — |
| case-05 | ✗→✓ | ▲ Improved | — | — |
| case-12 | ✗→✓ | ▲ Improved | — | — |
| case-01 | ✗→✓ | ▲ Improved | — | — |
Send a question or task to Codex and return the response.
Do not pass free-form user text to the shell unquoted. The question or task may contain spaces or shell metacharacters such as (, ), ;, #, *, or [.
If the user only supplied a question or task, execute:
bash"${CLAUDE_PLUGIN_ROOT}/scripts/ask-codex.sh" "$ARGUMENTS"
If the user supplied flags such as --codex-model or --codex-timeout, reconstruct the command so those flags remain separate shell arguments and the remaining free-form question is passed as one quoted final argument.
Example:
bash"${CLAUDE_PLUGIN_ROOT}/scripts/ask-codex.sh" --codex-model gpt-5.5:high "Review the following round summary (M4)..."
Never run this unsafe form:
bash"${CLAUDE_PLUGIN_ROOT}/scripts/ask-codex.sh" $ARGUMENTS
because the shell will re-parse the question text and can fail before ask-codex.sh starts.
| Exit Code | Meaning | |-----------|---------| | 0 | Success - Codex response is in stdout | | 1 | Validation error (missing codex, empty question, invalid flags) | | 124 | Timeout - suggest using --codex-timeout with a larger value | | Other | Codex process error - report the exit code and any stderr output |
.humanize/skill/<timestamp>/output.md for referencegpt-5.5:high with a 3600-second timeout| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 14 counted toward the lift figure. The other 8 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 14 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.