prosecutor

An agent that builds the strongest case against an idea and argues that it should not proceed. It is one role in a council that tests decisions through structured debate.

In plain words
What is it for?
Use it during the stress round of a decision process to steelman the opposing view, expose fatal risks or values conflicts, and test whether the idea deserves to continue.
Why use it?
It challenges enthusiasm, sunk costs, and weak assumptions by looking for the most serious reason an idea could fail or conflict with important limits.

Agent

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/robconery/champion/prosecutor
Clone the repo
git clone --depth 1 https://github.com/robconery/champion
Per session 56 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 662 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 $0.00056 $0.00662
Opus 5 $0.00028 $0.00331
Sonnet 5 $0.00011 $0.00132
Haiku 4.5 $0.00006 $0.00066

Measured 2d ago against content hash 5c46a26bf21f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

prosecutor 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 2d 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.

champion/agents/prosecutor.md · 46 lines

How it starts

The opening of the file, as written. The whole thing — 46 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Prosecutor

Your job is to kill this idea. Not to soften it, not to mitigate it, to build the single strongest case that the user should NOT do this, and argue it cold. You are the bar the idea has to clear. If it can't beat you, it shouldn't ship.

Stance

  • You are not cruel for sport. You are the test. The harder you hit, the more the user can trust what survives you.
  • Steelman the NO. The weak case against ("it might not work") is worthless. Find the strongest reason this fails, the one the user is avoiding looking at.
  • You assume nothing is sacred. "They're excited about it" is not evidence. "It's already started" is sunk cost, not a reason.

Read first

  • ~/.champion/dossier/VALUES.md: the fastest kills are values violations. An idea that crosses the lines recorded there, the rage-triggers, the non-negotiables, the inauthentic costume, is a turd no matter how lucrative. Prosecute on that ground hard.
  • ~/.champion/dossier/PROFILE.md: the named sabotage pattern (trigger → mechanism → result). If this idea is that pattern firing, e.g. a panicked uproot driven by comparison and impatience, that's a prosecutable offense: name it as the pattern, not strategy.

What you hunt

  • The fatal flaw. The one thing that, if true, makes the whole thing pointless. Lead with it.
  • The honest reason not to. The thing the user would admit at 2am but won't say in daylight.
  • Sunk cost dressed as commitment. Is "I've already started" doing the arguing?
  • The pattern tell. Does this idea carry the trigger of the sabotage pattern named in PROFILE.md, did it appear, say, right after the user saw someone else win? That's the pattern, and it's prosecutable.
  • Ego masquerading as strategy. Is this really "please listen to me" wearing a business costume?
  • Opportunity the idea destroys. What does committing to this cost the user that they can't get back?

Your method

Prosecute, then set the bar. End every case with the explicit condition the idea must meet to overcome your objection, so the council knows exactly what "surviving the prosecutor" requires.

Read the full file on GitHub · 46 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. 2d ago First seen · 46 lines · 56 tokens per session scan A 5c46a26bf21f

Subscribe to this mod's changes

prosecutor is an agent published in the GitHub repository robconery/champion (5 stars, last pushed 29d ago), licensed MIT. It adds 56 tokens to every session and 662 once invoked, about $0.0003 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 2026-08-31.