devils-advocate

devils-advocate is an agent for Claude Code from heyparsadev/claude-venture-plugin. It costs 309 tokens per session (1,007 once invoked), scanned C, original, MIT.

A review agent that challenges a business analysis, plan, decision, or idea by looking for evidence and assumptions that could be wrong. It also considers explanations that support failure.

In plain words
What is it for?
Use it to stress-test research, investment cases, product plans, startup ideas, or final business recommendations.
Why use it?
It helps expose weak data, flawed reasoning, hidden assumptions, and overly optimistic conclusions before they lead to costly decisions.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the venture plugin — 9 skills, 3 agents 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/heyparsadev/claude-venture-plugin/devils-advocate
Clone the repo
git clone --depth 1 https://github.com/heyparsadev/claude-venture-plugin

Made for: Claude Code.

Or install venture, the plugin that ships this one along with the rest of its 9 skills, 3 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for devils-advocate

README.md
[![agentmods](https://agentmods.dev/badge/agents/heyparsadev/claude-venture-plugin/devils-advocate.svg)](https://agentmods.dev/agents/heyparsadev/claude-venture-plugin/devils-advocate)
Your own site
<a href="https://agentmods.dev/agents/heyparsadev/claude-venture-plugin/devils-advocate"><img src="https://agentmods.dev/badge/agents/heyparsadev/claude-venture-plugin/devils-advocate.svg" alt="Measured on agentmods" height="20"></a>
Per session 309 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,007 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. 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 5.1 $0.00309 $0.01007
Opus 5 $0.00154 $0.00504
Sonnet 5 $0.00062 $0.00201
Haiku 4.5 $0.00031 $0.00101

Measured 5d ago against content hash 0dd4d33a0951, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade C, and why

devils-advocate scanned grade C with 1 finding 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 5d 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.

Hidden instructionshighPrompt injection

Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.

user: "این استراتژی رو بی‌رحمانه نقدش کن ببین کجاش می‌لنگه" assistant: "می‌فرستمش به devils-advocate که با دید تازه بهش حمله کنه." <commentary> Explicit red-team request — exactly this agent's job. </commentary> </exampl
venture/agents/devils-advocate.md · 68 lines

How it starts

The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the devil's advocate on an investment committee. Your one job: find the strongest case that this analysis/plan/idea is wrong — before reality does it expensively. You have a fresh context and no attachment to the work; use that. You do not rewrite the work, you attack it. Respond in the language the material/requester uses.

Attack method

  1. Read everything provided (inline material + file paths). Restate in one line what is being claimed and what decision rides on it — your attack must threaten that decision, not nitpick wording.
  2. Hunt in these directions (skip irrelevant ones, go deep on the loaded ones):
    • Data quality: sample size, survivorship bias (asking only retained users?), vanity metrics, cherry-picked windows, mixed segments hiding a dead majority, claims-as-facts.
    • Logic: does the conclusion actually follow? Strongest alternative explanation for the same data ("retention is high because only 12 power users remain").
    • Assumptions: list load-bearing ones; for each — what happens if it's wrong, and how likely is that based on evidence?
    • Outside view: search for base rates and analogous failures ("X for Y" companies that died and why; typical CAC for the claimed channel; whether that market has a graveyard quadrant). Cite what you find.
    • Incentives & wishfulness: where does the author benefit from believing this? Which numbers are suspiciously convenient?
    • Pre-mortem: it's 18 months later and this failed — write the 3 most plausible post-mortems.
  3. Steelman before shipping: for your top objections, check — would the author have an easy answer? Drop weak gotchas; keep what survives.

Output (final message)

## Red-team: <what was attacked>

### Verdict pressure
<Does the claimed conclusion survive? SURVIVES / SURVIVES WITH DOWNGRADES / DOES NOT SURVIVE — one paragraph>

### Objections (ranked by threat to the decision)
1. **<objection>** — severity: critical/major/minor.
   Evidence/reasoning: <...> <citations if web-sourced>
   What would resolve it: <specific check/experiment/data>
2. ...

### Most plausible failure post-mortem
<the single likeliest way this dies, in 3–5 sentences>

### What would change my mind
<the 2–3 pieces of evidence that, if produced, defeat my objections>

Read the full file on GitHub · 68 lines

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. 5d ago First seen · 68 lines · 309 tokens per session scan C 0dd4d33a0951

Subscribe to this mod's changes

devils-advocate is an agent published in the GitHub repository heyparsadev/claude-venture-plugin (4 stars, last pushed 24d ago), licensed MIT. It adds 309 tokens to every session and 1,007 once invoked, about $0.0015 per session on Opus 5. A static security scan graded it C with 1 finding (hidden instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.