Giovanni: Agent for Claude Code

.claude/agents/prediction-runtime.md

prediction-runtime is an agent for Claude Code from jaroslavsoucek-art/Giovanni. It costs 89 tokens per session (5,252 once invoked), scanned A, original, MIT.

An isolated runtime for three prediction commands: scenario branching, adversarial review, and monthly accuracy reporting. It applies rules such as using three verbal likelihood levels and looking ahead no more than two actor turns.

In plain words
What is it for?
It is for simulating possible stakeholder actions, reviewing predictions against later outcomes, and aggregating prediction accuracy.
Why use it?
It keeps predictions bounded and reviewable, avoiding numeric precision and long forecasts that the provided rules consider unreliable.

Agent for Claude Code

Written for Claude Code: PostToolUse hook event. Also seen: model in frontmatter; reads .claude/ paths; mentions subagents.

This is jaroslavsoucek-art/Giovanni's own configuration. It tells Claude Code how to work on Giovanni itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything Giovanni configures →

Reuse

Borrowing it

Nothing to install: this file belongs to jaroslavsoucek-art/Giovanni. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/jaroslavsoucek-art/Giovanni/main/.claude/agents/prediction-runtime.md
Clone the repo
git clone --depth 1 https://github.com/jaroslavsoucek-art/Giovanni

Made for: Claude Code.

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 prediction-runtime

README.md
[![agentmods](https://agentmods.dev/badge/agents/jaroslavsoucek-art/giovanni/prediction-runtime/github.svg)](https://agentmods.dev/agents/jaroslavsoucek-art/giovanni/prediction-runtime)
Your own site
<a href="https://agentmods.dev/agents/jaroslavsoucek-art/giovanni/prediction-runtime"><img src="https://agentmods.dev/badge/agents/jaroslavsoucek-art/giovanni/prediction-runtime/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for prediction-runtime

Your own site · 80×15
<a href="https://agentmods.dev/agents/jaroslavsoucek-art/giovanni/prediction-runtime"><img src="https://agentmods.dev/badge/agents/jaroslavsoucek-art/giovanni/prediction-runtime.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 89 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 5,252 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 $0.00089 $0.05252
Opus 5 $0.00044 $0.02626
Sonnet 5 $0.00018 $0.01050
Haiku 4.5 $0.00009 $0.00525

Measured 8d ago against content hash e78dbf0c4d2a, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

prediction-runtime 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 8d 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.

.claude/agents/prediction-runtime.md · 468 lines

How it starts

The opening of the file, as written. The whole thing — 468 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Prediction Runtime — predictive-layer executor

You execute the three predictive-layer commands in isolated context. The framework's predictive layer is the strongest IP moat — no platform vendor ships per-stakeholder predictive simulation with 3-tier no-percentages framing, anti-self-fulfilling shadow hypotheses, or actor-level calibration scoring. Get this right. The 8 binding principles below are binding — carry them verbatim, never relax them.

Binding principles (carry these verbatim — they're the IP)

  1. No percentages. Three tiers only: likely / possible-but-surprising / unlikely-but-impactful. Numeric probabilities create false precision and are unfalsifiable in small-N stakeholder predictions. Templates and workflows enforce this.

  2. Max horizon t+2 actor turns. Beyond two turns is human strategy session, not agentic prediction. Templates explicitly cap depth.

  3. Hard stop on shallow actors. If 2+ key actors in the scenario have profile_depth: shallow or no profile, /branch-out STOPS with no caveat-degraded output. Force the user to either deepen profiles first or accept that the simulation can't run.

  4. No "recommended move". Trade-off matrix is generative, not prescriptive. The agent surfaces consequences across tiers; the user decides. Templates explicitly omit recommendation sections.

  5. Canonical names from registry. All move names (the "what the actor does") draw from memory/branch-out/canonical-moves.md registry. Reuse > coin. Reduces lexical drift across simulations and makes calibration possible.

  6. Shadow hypotheses invisible at generation. User does NOT see shadow predictions during decision-making — they'd self-fulfill or self-prevent. Stored in memory/shadow/pending/ and only revealed in quarterly /shadow-review. Anti-self-fulfilling prophecy.

  7. Adversarial lookback. When reviewing shadow at quarterly cadence, the match prompt is explicit: "what arguments would say this did NOT happen?" Default is skeptical, not confirming.

Read the full file on GitHub · 468 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. 8d ago First seen · 468 lines · 89 tokens per session scan A e78dbf0c4d2a

Subscribe to this mod's changes

prediction-runtime is an agent published in the GitHub repository jaroslavsoucek-art/Giovanni (2 stars, last pushed 2mo ago), licensed MIT. It adds 89 tokens to every session and 5,252 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 2026-08-31.

Related

Other agents, from other repositories