Skip to content
SkillModel-led986 tokens in use

second-opinion

Runs an external CLI review (Codex, Antigravity or Gemini) on your uncommitted changes, a branch diff or a commit, and reports the findings with severity and file lines.

8,320 stars on the repository CC-BY-SA-4.0trailofbits/skills · plugins/second-opinion/skills/second-opinion
/plugin marketplace add trailofbits/skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing second-opinion costs

63tokens idle
986tokens in use
4,137tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

shelldeclared

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter12/15
  • Structure8/25
  • Examples3/15
  • Output10/15
  • Size15/15
  • Resources12/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity100
  • Maintenance52
  • Community64
  • Maturity32
  • Momentum40

second-opinion asks another AI tool to review your code. It captures one patch (uncommitted work including untracked files, a branch against a base, or a single commit), builds a review prompt with your focus and project conventions, and sends it to the Codex CLI, the Antigravity CLI, the Gemini CLI, or two of them for a comparison.

The skill reports findings only. It applies no fixes and posts nothing to a remote service, and a failed or empty run is reported as an incomplete review, never as a clean one.

What You Get

  • Findings from an independent reviewer, sorted by severity, with file and line references
  • A side-by-side summary of agreements and disagreements when two providers review the same diff
  • A clear report of failures, coverage gaps and partial output when a CLI is missing, unauthenticated or times out
  • The same captured patch used for every provider, including untracked files for uncommitted reviews

Common Use Cases

  • Getting a Codex review of work in progress before committing
  • Comparing Codex and Antigravity on a branch against origin/main
  • Asking Gemini CLI to check a specific commit for security issues

What Makes It Unique

It treats the external reviewer as a second reader, not as proof: agreement between providers is not counted as confirmation, and incomplete runs are never presented as clean reviews. Codex runs in a read-only sandbox with a JSON schema for findings, and the skill refuses to enable automatic write approval, slash commands in the diff or a skipped trust prompt to force a run through.

Under The Hood

The skill is an instruction file with reference notes for each provider. It runs git commands to capture the patch, creates temporary files with mktemp outside the checkout, and writes a prompt file that contains your diff and project instructions. It then runs the codex, agy or gemini command line tool, which sends that prompt and your code to the provider’s service over the network, using your own account. Output and diagnostics go to temporary files that the skill reads back. The Codex run uses a read-only sandbox. If a tool is missing, the skill reports the install command from its notes and does not install it. It does not edit your source files.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers