agentleFS
Sign inStart free
skyremoteskyremote/chief-of-staff-kit/skills/advisor/SKILL.md

advisor

Skill22 starsChanged 18 days agoMITView on GitHub

Adopt and verify Anthropic's advisor primitive for an installed crew. Checks Claude Code support, configures advisorModel without overwriting settings, adds tightly scoped escalation discipline to high-judgement agents and verifies the result.

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

11 rules. 0 are common, 7 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 21 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run claude --version and confirm that the installed release exposes /advisor or the advisorModel setting.

    line 25
  2. Distinctive:

    Read the current main model without exposing other settings.

    line 27
  3. Distinctive:

    Ask which advisor model to use if the user has not specified one.

    line 28
  4. Very specific:

    Create a timestamped backup of ~/.claude/settings.json.

    line 30
  5. Very specific:

    Add a short Reaching for the advisor section only to high-judgement parts of the crew: in the one-loop shape, the chief-of-staff main loop, …

    line 32
  6. Distinctive:

    If the same crew is installed for Codex, do not add Anthropic advisor instructions to Codex agents.

    line 37
  7. Very specific:

    Start a fresh Claude Code session and run /advisor to verify the primitive is available.

    line 45
  8. Distinctive:

    If live invocation is unavailable, distinguish a configuration success from an unverified runtime result.

    line 46
  9. Distinctive:

    before committing to a consequential approach where judgement changes the outcome.

    line 51
  10. Distinctive:

    after the same failure repeats twice.

    line 52
  11. Distinctive:

    before declaring high-impact work complete.

    line 53

“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

Adopt the advisor for the crew

The advisor primitive lets a working Claude model consult a stronger model inside the same session. It is not another subagent and does not change the crew's shape.

Guardrails

  • Inspect the installed Claude Code version and existing settings before changing anything.
  • Back up ~/.claude/settings.json; it may contain secrets and permission rules. Never print or commit its contents.
  • Preserve every existing setting. Add or update only advisorModel.
  • The advisor must be at least as capable as the main model. If the desired pairing is uncertain or unsupported, stop and explain the pairing instead of silently disabling it.
  • Advisor calls consume plan/API quota. Reserve them for judgement points, repeated failure and final independent review; do not add advisor instructions to every low-judgement role.

Runbook

  1. Run claude --version and confirm that the installed release exposes /advisor or the advisorModel setting. If it does not, recommend updating Claude Code and stop.
  2. Read the current main model without exposing other settings.
  3. Ask which advisor model to use if the user has not specified one. Recommend the strongest model available within their quota, with the main/advisor compatibility explained.
  4. Create a timestamped backup of ~/.claude/settings.json.
  5. Parse the JSON, change only advisorModel, and write valid formatted JSON.
  6. Add a short Reaching for the advisor section only to high-judgement parts of the crew: in the one-loop shape, the chief-of-staff main loop, the code-reviewer agent and the architecture, problem-solving and web-build skills; in the legacy shape, the orchestrator, architect, code auditor, problem solver and web builder. Keep any existing tailored version instead of duplicating it.
  7. If the same crew is installed for Codex, do not add Anthropic advisor instructions to Codex agents; Codex uses its own reviewer/subagent mechanisms.

Verification

  • Parse the settings JSON.
  • Report the advisor model value without printing unrelated settings.
  • Confirm which agent files contain the escalation discipline.
  • Start a fresh Claude Code session and run /advisor to verify the primitive is available. If live invocation is unavailable, distinguish a configuration success from an unverified runtime result.

The discipline block should say to consult the advisor only:

  • before committing to a consequential approach where judgement changes the outcome;
  • after the same failure repeats twice;
  • before declaring high-impact work complete.

On routine or mechanical work, do not consult it.

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.

Check your setup, free