Skip to content
Back to skills

Gemini Result Handling

ASecurity

Internal guidance for presenting Gemini helper output back to the user

  • 4 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added May 27, 2026
development

Works with

  • claude code

Security analysis

A100/100

Scanned May 27, 2026

npx -y skills add taibaran/gemini-plugin-cc --skill gemini-result-handling --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Gemini Result Handling?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Gemini Result Handling
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/taibaran-gemini-result-handling/badge)](https://www.skillsdirectory.com/skills/taibaran-gemini-result-handling)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: gemini-result-handling
description: Internal guidance for presenting Gemini helper output back to the user
user-invocable: false
---

# Presenting Gemini output

Use this skill from the main Claude Code thread when surfacing output from a Gemini companion command (review, adversarial-review, task, result).

## Default rule: verbatim

Return Gemini's stdout to the user **verbatim**. Do not paraphrase, summarize, re-style, or "improve" the output. The user invoked Gemini deliberately to get a second opinion in Gemini's voice — sanitizing it through Claude defeats the purpose.

## When to add framing

Wrap Gemini's output in a short header **only** when:
- It came from `/gemini:rescue` and the user might not realize a delegation just happened. A one-line "Gemini's response:" header is fine.
- It came from `/gemini:review` and arrived inside an `AskUserQuestion` flow where the user picked "background". A one-line follow-up "Gemini review (job <id>) completed — output above" is fine.

## When NOT to add framing

- `/gemini:ask` — the user asked a direct question, just return the answer.
- `/gemini:status` — the companion already renders a Markdown table; pass it through.
- `/gemini:result` — the output is already structured. Don't re-format.

## Auth-error hints

If the companion's stderr contains a `[hint: ...]` line, surface that hint to the user and suggest `/gemini:setup`. Do not try to recover yourself.

## JSON review output

If `/gemini:review --json` was used, the companion strips Gemini's wrapper (`{session_id, response, stats}`), removes markdown code fences, and emits clean JSON. The shape is checked against `schemas/review-output.schema.json` at the top level (verdict / summary / findings / next_steps must be present and verdict must be in the enum); deeper field-level validation is the caller's responsibility. When forwarding it:
- If the user asked for JSON specifically, pass it through unmodified.
- If the user did not ask for JSON, render the JSON into a compact Markdown report grouped by `severity`. Preserve `file`, `line_start`, `line_end`, and `recommendation` for each finding.
- If the companion's stderr contains `schema mismatch`, surface that warning along with the raw output — Gemini may have ignored the schema instruction.

## Background jobs

When `/gemini:rescue` is launched in background mode, the companion writes to a log file and reports the job ID. Tell the user to run `/gemini:status` or `/gemini:result <id>` later — do not poll on their behalf.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.