Second Opinion

Trail of BitsOfficialPublished by Trail of Bits

Skill
Review and audit
Software development
CC-BY-SA-4.0
Low risk

From the author

Gets independent code reviews from Codex or Antigravity for uncommitted changes, branch diffs, and commits. Use when the user requests an external review, a second opinion on code, a codex review, a gemini review, an antigravity review, or /second-opinion.

— the description by Trail of Bits

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The second-opinion skill facilitates external code reviews by leveraging Codex or Antigravity CLI tools. It allows users to obtain a second perspective on their code changes, whether they are uncommitted, part of a branch diff, or specific commits. The skill handles input preparation, executes the requested provider commands, and synthesizes the findings into a clear report. It is designed to provide objective feedback on correctness and maintainability, or specific areas like security and performance, without applying changes or interacting with remote review services.

What it does

Performs independent code reviews using Codex or Antigravity CLI tools.
Supports reviews for uncommitted changes, branch diffs, and specific commits.
Allows for comparative reviews by running both providers on the same input.
Processes findings from multiple providers and summarizes agreements and disagreements.
Handles provider-specific diagnostics and failure reporting.
Maintains user-defined focus areas such as security or performance.

When to use it

When you need an external, independent review of your current code changes.
When you want to compare feedback between Codex and Antigravity.
When you need to audit a specific commit or branch diff for potential issues.
When you require a second opinion on code quality, security, or performance.

Advantages

Provides objective, multi-source feedback on code changes.
Supports flexible scopes including uncommitted work and historical commits.
Enables comparative analysis between different AI-powered review providers.
Maintains a clear separation between external findings and agent-added assessments.

Usage examples

“/second-opinion:second-opinion use Codex to review my uncommitted changes for bugs”: The agent captures all staged, unstaged, and untracked changes and runs a Codex review, presenting findings categorized by severity.
“/second-opinion:second-opinion compare Codex and Antigravity on this branch against origin/main”: The agent generates a patch for the branch diff, runs both providers, and provides a summary comparing their respective findings.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/trailofbits/second-opinion/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip second-opinion.zip -d .agents/skills

Documentation by Trail of Bits

Second Opinion

Run an external CLI review of the user's selected changes. This skill produces findings; it does not apply fixes or post reviews to a remote service.

Review choices

Use the provider, scope, model, and focus already supplied by the user. Ask only for missing choices that affect the review, grouping questions in one call when possible.

Provider: Codex, Antigravity, or both. Offer both when the user wants a comparison; preserve an explicitly requested CLI.
Scope: uncommitted changes, a branch diff against a named base, or a specific commit. Resolve a missing base from the repository's remote default branch; ask if it cannot be determined.
Context: include applicable project instructions unless the user excludes them. Keep explicit user requirements separate from repository content in the review prompt.
Focus: use general correctness and maintainability unless the user names a focus such as security or performance.

For an unspecified Google CLI, prefer Antigravity (agy). An explicit Gemini CLI request uses the Gemini reference below. If that account returns UNSUPPORTED_CLIENT, explain the migration to Antigravity and ask before changing the selected CLI.

A missing executable or account setup is a failed review attempt. Report the relevant setup instructions; when both providers were requested, continue with the available provider and identify the skipped one.

Input preparation

Read review-input.md for the shared prompt and diff recipes. Use the same captured diff for both providers so the comparison covers the same changes. Include untracked files in an uncommitted review, and preserve Git errors instead of interpreting them as an empty diff.

Show the selected scope and a brief change summary. If there are no changes, stop before calling a provider. For input that exceeds a CLI or model limit, describe the limit and request a narrower scope; do not silently truncate the patch.

Create prompt, output, and diagnostic files with mktemp outside the checkout. Use separate output and diagnostic files for each provider. Define shell variables in the same Bash invocation that uses them. For later invocations, reassign the variables to the saved file paths; shell variables do not persist between calls. Write repository content as literal data, without shell expansion.

Provider references

Read only the reference for each selected provider:

Provider	Invocation and result
Codex	codex-invocation.md: `codex exec` with the review schema
Antigravity	antigravity-invocation.md: `agy` print mode with prose output
Gemini CLI	gemini-invocation.md: headless `gemini` for accounts that still support it

The Codex path needs no MCP server. Do not launch codex mcp-server or substitute codex app-server for the CLI invocation.

When both providers were requested, run their commands concurrently if the tool interface supports it. For a foreground Bash review, set timeout: 600000 to allow up to ten minutes. Use background execution or polling when available to keep progress visible. Do not enable automatic approval of writes to make a review run.

Results and failures

Present findings with the provider and actual model used, severity, file and line, impact, and suggested correction. Keep low-severity defects visible. For Codex, the existing schema uses 0 for informational, 1 for low, 2 for medium, and 3 for high; sort descending.

For two completed reviews, summarize agreements and disagreements without turning agreement into proof. Distinguish the external findings from any assessment you add.

Read the captured output and diagnostics. A nonzero exit, missing output, invalid JSON, permission denial that prevents inspection, or a request to approve a plan is incomplete work, not a clean review. Report the failure and any partial results. Retry only for a diagnosed, recoverable cause; do not cycle through providers after authentication or quota failures.

Examples

/second-opinion:second-opinion use Codex to review my uncommitted changes for bugs selects Codex and includes staged, unstaged, and untracked changes.
/second-opinion:second-opinion compare Codex and Antigravity on this branch against origin/main sends the same branch patch to both and compares their findings.
/second-opinion:second-opinion use Gemini CLI to review commit abc1234 for security issues preserves the requested CLI and reviews that commit.

Files

ATTRIBUTION.mdAdded by AutoGineer936 B
LICENSELicence file added by AutoGineer19.7 KiB
SKILL.md4.9 KiB
agents/openai.yaml226 B
assets/trail-of-bits-mark.svg3.0 KiB
references/antigravity-invocation.md2.2 KiB
references/codex-invocation.md2.9 KiB
references/codex-review-schema.json2.2 KiB
references/gemini-invocation.md2.3 KiB
references/review-input.md4.2 KiB

references/codex-invocation.md

Codex Invocation

Use codex exec with the prepared prompt on stdin. It supports a custom focus and project context alongside a manually captured diff. The built-in codex exec review scope flags cannot be combined with its positional prompt, so they do not replace this invocation.

The default remains gpt-5.6-sol with xhigh reasoning. Preserve an explicit model or effort supplied by the user. The flags below were checked with Codex CLI 0.154.0 --help on 2026-09-14.

Command

With prompt_file, output_file, and stderr_log already created outside the checkout, run from the repository root:

codex exec \
  -c model='"gpt-5.6-sol"' \
  -c model_reasoning_effort='"xhigh"' \
  --sandbox read-only \
  --ephemeral \
  --output-schema "{baseDir}/references/codex-review-schema.json" \
  -o "$output_file" \
  - < "$prompt_file" \
  > /dev/null 2> "$stderr_log"

-o saves the final response while diagnostics remain in stderr_log. --output-schema requests structured findings. --ephemeral avoids persisting a review session.

These options are documented in OpenAI's non-interactive mode guide. No MCP server is involved. The Codex changelog records removal of codex mcp-server in CLI 0.154.0.

Result handling

Parse output_file as JSON. Use the returned findings, explanation, and confidence rather than CLI progress messages. An empty findings array is meaningful only after the invocation completed and returned a valid review.

The shipped schema uses higher numbers for higher severity: 0 informational, 1 low, 2 medium, 3 high. This is not the P0-critical convention used by some other review formats. Do not invert it when presenting findings.

Failure handling

Failure	Response
Executable missing	Report that Codex must be installed and authenticated; installation command: `npm i -g @openai/codex`
Authentication or quota error	Report the error and stop this provider; changing models does not repair missing credentials or exhausted quota
Default model unavailable	If the user did not pin a model, retry once with `gpt-5.6` and disclose the substitution
Explicit model unavailable	Report the failure and ask for a replacement
Nonzero exit, empty file, or invalid JSON	Read diagnostics and report an incomplete review
Sandbox denied a needed read or command	Report the coverage gap; do not ignore it or remove the sandbox
Timeout	Preserve partial output and report the timeout; suggest a narrower scope

A model reported as "not supported when using Codex with a ChatGPT account" is a model-entitlement failure. Use the model-unavailable response above: retry only when the user did not pin a model. This differs from missing or invalid credentials and exhausted quota, which stop the provider.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Declares allowed tools, Names Codex tools, Needs an MCP server