agentleFS
jsvillalbatjsvillalbat/sevito-skills/skills/advisor/SKILL.md

advisor

Skill0 starsChanged 3 months agoMITView on GitHub

Gets an independent second opinion from a different-provider model — not the one currently working — on a plan, a diff, or an open question, without gating or auto-deciding anything. Use before committing to a risky or expensive plan, when stuck on a hard technical decision, or whenever a genuinely different reasoning process would help sanity-check the current approach.

21 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use orchestrator and ship.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • tesla0225Skill

    orchestrator

    Issue/要件を受けたら分解→担当割当→実装→検証→統合を回す手順。未整備なら /repo-init でbootstrap。

    Also in 3 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 4 projects that use this

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 21 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before committing to a plan that's expensive or hard to reverse.

    line 24
  2. Distinctive:

    Not for routine decisions — reserve it for calls where being wrong is costly enough that the extra step earns its keep.

    line 28
  3. Very specific:

    Check what's available: Look for a different-provider CLI (which codex, which gemini, or similar).

    line 39
  4. Distinctive:

    Invoke it and read the response critically: Don't adopt the advisor's opinion uncritically, and don't dismiss it because it's inconvenient.

    line 51
  5. Distinctive:

    Don't average the opinions — report the actual disagreement.

    line 68
  6. Distinctive:

    This is a routine, low-stakes decision: Don't reach for this — advisor exists for calls where being wrong is expensive, not as a rubber …

    line 69

“Common” is measured: kept by at least 6 in 10 of the 21 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Advisor

A second opinion from the same model family — even a fresh context — tends to agree with the first opinion more often than a genuinely independent reasoning process would. This skill gets an actual second opinion: a different model provider, asked a precise question, with no obligation to follow its answer.

This is the lightweight, on-demand cousin of codex-judge: codex-judge is a formal, rubric-scored gate for a post-review completion decision; advisor is for anything else — a plan, a hard call, an open question — where you want a genuine outside view but nothing should be auto-approved or auto-rejected.

When to use this

  • Before committing to a plan that's expensive or hard to reverse.
  • You're stuck between two approaches and want a reasoning process that isn't anchored on the same assumptions you already have.
  • The user explicitly asks for "a second opinion" or "what would another model say."
  • Not for routine decisions — reserve it for calls where being wrong is costly enough that the extra step earns its keep.

Instructions

  1. Write a precise, self-contained brief. The advisor gets nothing from this conversation except what you give it. Include: the goal, the relevant constraints, the actual plan/diff/question, and a specific question — "what am I missing", "is there a simpler approach", "what would break under load" — not a vague "thoughts?". A vague brief gets a vague, unhelpful answer back. See references/brief-template.md.
  2. Check what's available. Look for a different-provider CLI (which codex, which gemini, or similar). It must be a genuinely different model family than whatever produced the plan/diff being reviewed — a second instance of the same model isn't independent, however fresh its context.
  3. If nothing is available, degrade honestly — don't skip silently. Say plainly that no second-provider access exists. Do one of:
    • a deliberate cold review yourself (see verify-before-done's references/independent-review.md for the cold-read protocol), or
    • hand the user the brief you wrote and let them paste it into another tool themselves.
  4. Invoke it and read the response critically. Don't adopt the advisor's opinion uncritically, and don't dismiss it because it's inconvenient. If it disagrees with the current plan, that disagreement is the useful part — surface it explicitly to the user rather than quietly picking a side.
  5. Report the opinion and the reconciliation, not just "got a second opinion." State what the advisor said, where it agrees or disagrees with the current approach, and what you're doing as a result (adopting its suggestion, keeping the original approach with a stated reason, or flagging the disagreement for the user to decide).

Edge cases

  • The advisor's answer is generic or unhelpful. The brief probably wasn't specific enough — tighten the question and try again rather than reporting a non-answer as "consulted, no issues found."
  • The advisor and the reviewer/judge disagree with each other, not just with you. Don't average the opinions — report the actual disagreement; it's signal, not noise.
  • This is a routine, low-stakes decision. Don't reach for this — advisor exists for calls where being wrong is expensive, not as a rubber stamp for every plan.

Reference files

  • references/brief-template.md — a fill-in-the-blanks template for writing a precise, self-contained brief for the advisor.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.