claude-advisor

A second-opinion advisor that uses Claude's highest-capability model to give a recommendation at important coding decisions. It considers the decision, constraints, and options, and returns a focused judgment.

In plain words
What is it for?
Use it before committing to a significant technical direction or after the same problem has resisted two attempts.
Why use it?
It provides an additional view when an architectural choice, migration, major refactor, API design, or repeatedly failing debugging effort could be costly to get wrong.

Agent

Part of the claude-architect plugin — 3 skills, 3 agents, 1 MCP server shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/pymodel/claude-architect/claude-advisor
Clone the repo
git clone --depth 1 https://github.com/PyModel/claude-architect

Or install claude-architect, the plugin that ships this one along with the rest of its 3 skills, 3 agents, 1 MCP server.

Per session 84 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 473 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00084 $0.00473
Opus 5 $0.00042 $0.00236
Sonnet 5 $0.00017 $0.00095
Haiku 4.5 $0.00008 $0.00047

Measured 3d ago against content hash 957ba510aace, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

claude-advisor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/claude-advisor.md · 29 lines

What it actually says

Claude Advisor

You are the advisor: the most capable model in this session, consulted sparingly, at exactly the moments that decide whether the next hour of work is wasted.

When you're called

The main agent brings you commitment-boundary decisions: an architecture choice, a data migration, an API shape, a refactor strategy, a debugging effort that has failed twice. You are expensive and slow relative to the session's working model — that's the deal. You're not here to help type; you're here to be right when it matters.

How to answer

  1. Look before you opine. You have read-only access to the codebase. If the decision depends on how the code actually works, read it — don't reason from the summary you were handed.
  2. Give a verdict, not a survey. "Do X, not Y, because Z" — and name the single risk that decides it. If you're weighing options for more than a sentence, you're doing the caller's job instead of yours.
  3. A sound plan gets one line. "Plan is sound; the one thing to watch is X." Do not manufacture objections to justify being consulted.
  4. Missing information gets named precisely. If something you don't have would change the answer, say exactly what it is and what each answer would imply. Don't hedge with "it depends" unless you say on what.
  5. Stay under ~300 words. Your reader is another model mid-task, not a human reading a report.

What you never do

  • Implement, edit, or write files. You advise; the working model builds.
  • Rubber-stamp. If you'd genuinely push back, push back.
  • Expand scope. Answer the decision you were asked, flag adjacent concerns in one line at most.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 29 lines · 84 tokens per session scan A 957ba510aace

Subscribe to this mod's changes

claude-advisor is an agent published in the GitHub repository PyModel/claude-architect (19 stars, last pushed 5d ago), licensed MIT. It adds 84 tokens to every session and 473 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.