What makes Prisoner's Dilemma notable?

1.1K downloads. Trust evidence available

How should Prisoner's Dilemma be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Prisoner's Dilemma

Activate when: user asks 'why does everyone keep doing X when it's obviously bad for all of us', 'how do we get out of this race to the bottom', 'should we t... Skill: Prisoner's Dilemma Owner: deciqai Summary: Activate when: user asks 'why does everyone keep doing X when it's obviously bad for all of us', 'how do we get out of this race to the bottom', 'should we t... Tags: latest:1.0.5 Version history: v1.0.5 | 2026-07-16T18:11:49.118Z | user Description tail link + agents machine-readable metadata line (deciqai.com/s/prisoners-dilemma.json) v1.0.4 | 2026-07-09T11:21:00.99

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.1k

Updated

Oct 11, 2026

Version

1.0.5

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.1K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1.1K downloadsadoption · observed Oct 11, 2026
Latest release: 1.0.5release · observed Jul 16, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17a4mqcnk515kvaca5ze55d0x88pfpx:prisoners-dilemma

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-deciqai-prisoners-dilemma/snapshot"

Documentation

CLAWHUB

143,406 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: prisoners-dilemma
description: "Activate when: user asks 'why does everyone keep doing X when it's obviously bad for all of us', 'how do we get out of this race to the bottom', 'should we trust them not to defect', 'we'd both be better off cooperating but it never happens', 'is this a prisoner's dilemma / Nash equilibrium / tragedy of the commons / race to the bottom'.
Do NOT activate when: the situation is zero-sum (one side's gain is the other's loss) — use a different game model; or when the parties already have a long-established repeated relationship with observable moves — use repeated-games-reputation instead. More: deciqai.com/c/prisoners-dilemma"
---

# Prisoner's Dilemma

## Overview

Whatever the other party does, **each player is individually better off defecting** — so both defect, and both end up worse than mutual cooperation. This is the structural skeleton beneath price wars, arms races, overfishing, and ad spend spirals. The problem is never character; it is structure. Exhortations to cooperate fail. Change the matrix.

Composes with: `second-order-thinking` for matrix redesign · `expected-value-and-kelly` for probabilistic payoffs · `repeated-games-reputation` for the iterated-game case.

## When to Use

- Two or more parties **would each do better cooperating**, but cooperation keeps failing to materialize
- Situation involves **price competition, capacity races, advertising arms races, or commons-style resource depletion**
- You are about to **negotiate or enter a partnership** and want to know whether the structure makes betrayal individually rational
- Someone asks directly about "prisoner's dilemma," "tragedy of the commons," "race to the bottom," or "Nash equilibrium"
- A present-day competitive sprint is in play — an **AI capex / compute arms race, AI-safety release race, or AI-native land-grab** where every player feels forced to move fast despite preferring collective restraint

**When NOT to use:** zero-sum games · pure coordination problems (Schelling) · long transparent repeated game with established reputations (use `repeated-games-reputation`) · low-stakes reversible decisions

## Coaching Novices (Adaptive Front Door)

- **Engine mode:** user has a concrete case → run The Process directly.
- **Coach mode:** user is unfamiliar or has no concrete case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-line what-it-is (≤2 sentences): some situations look like "people being stubborn" but are actually a structural trap — given the rules, defecting is individually rational, and exhorting people to cooperate won't work; you have to change the structure.
2. Check fit against When to Use / When NOT to use. If it's zero-sum or pure coordination, point elsewhere.
3. Elicit their real situation. Get a concrete case (a partnership, a market, a negotiation). > **[WAIT — do not advance until user responds]**
4. Run The Pr

_meta.json

{
  "ownerId": "kn754b8sk22s8c6gjxt02bftbn88q7ye",
  "slug": "prisoners-dilemma",
  "version": "1.0.5",
  "publishedAt": 1784225509118
}

references/sources.md

# Sources — prisoners-dilemma

> *Primary sources for the [prisoners-dilemma](../SKILL.md) skill.*

- Flood, M. M. (1952/1958). *Some Experimental Games*. RAND Research Memorandum RM-789-1; reprinted in *Management Science*, 5(1), pp. 5–26. The primary-source documentation of the first PD experiment, including the Alchian–Williams 100-round play and verbatim subject commentary. https://doi.org/10.1287/mnsc.5.1.5
- Tucker, A. W. (1980). "On Jargon: The Prisoner's Dilemma." *UMAP Journal*, 1, p. 101. Tucker's own retrospective account of inventing the two-prisoners exposition at Stanford in May 1950. Reprinted in *The Two-Year College Mathematics Journal*, 14(4), p. 326, 1983. https://doi.org/10.2307/3027101
- Poundstone, W. (1992). *Prisoner's Dilemma: John von Neumann, Game Theory, and the Puzzle of the Bomb*. Doubleday. The standard popular history; chapters 6–8 cover the Flood-Dresher-Tucker origin and contain extensive verbatim quotation of the original RAND notebooks. ISBN 978-0385415804.
- Axelrod, R. (1984). *The Evolution of Cooperation*. Basic Books. The foundational study of the iterated PD: the computer tournaments won by tit-for-tat, and chapter 4's analysis of the WWI "live and let live" trench system as a real-world iterated PD. ISBN 978-0465021215.
- Ashworth, T. (1980). *Trench Warfare 1914–1918: The Live and Let Live System*. Macmillan. The primary historical reconstruction — from diaries, letters, and unit histories — of tacit cooperation between front-line enemies on the Western Front; the empirical base for Axelrod's chapter 4.
- Von Neumann, J., & Morgenstern, O. (1944). *Theory of Games and Economic Behavior*. Princeton University Press. The founding text of game theory; PD-shaped problems are the canonical example of where Von Neumann's zero-sum apparatus stops giving useful answers and a richer framework is needed.
- Nash, J. F. (1950). "Equilibrium Points in n-Person Games." *Proceedings of the National Academy of Sciences*, 36(1), pp. 48–49. The equilibrium concept under which "both defect" is the predicted outcome of the one-shot PD. https://doi.org/10.1073/pnas.36.1.48
- Hardin, G. (1968). "The Tragedy of the Commons." *Science*, 162(3859), pp. 1243–1248. The canonical n-player PD generalization. https://doi.org/10.1126/science.162.3859.1243
- Ostrom, E. (1990). *Governing the Commons: The Evolution of Institutions for Collective Action*. Cambridge University Press. The empirical documentation of real-world communities that successfully escape commons-PDs without privatization or top-down regulation; the source of the 8 design principles for self-governed commons. ISBN 978-0521405997.
- Future of Life Institute (2023). "Pause Giant AI Experiments: An Open Letter." March 2023. A voluntary, multi-signatory call to pause training of systems more powerful than GPT-4 for at least six months; no lab paused — a real-world illustration that exhortation cannot move a dominant strategy. https://futureoflife.org/open-letter/pause-

examples/ai-safety-race-2023-2026.md

# Method in Action: The AI-Lab Safety Race (2023–2026)

> *Example for the [prisoners-dilemma](../SKILL.md) skill.*

Between the public launch of ChatGPT in late 2022 and 2026, the leading AI labs — OpenAI, Google DeepMind, Anthropic, Meta, and xAI, plus a fast-following Chinese cohort including DeepSeek — entered a period of extraordinarily fast, capital-intensive competition. Every lab publicly professes a commitment to safety; several were founded explicitly on it. Yet the observed equilibrium has been one of accelerating release cadence, escalating compute spend, and repeated compression of pre-release evaluation time. That gap — between what each lab says it prefers (careful, deliberate deployment) and what the field collectively produces (a sprint) — is the signature of a Prisoner's Dilemma. The problem is not that any lab is run by reckless people; it is that the structure rewards moving fast whatever the others do. Below, the case is walked through this skill's Process.

## 1. State the players and choices

**Players:** the frontier AI labs, reducible for diagnosis to two representative players — "Lab A" and "Lab B" (the same logic scales to n players and to the US–China framing below).

**Choices:** each lab can **Restrain** (cooperate — invest more in evaluations, red-teaming, and staged rollout; ship later) or **Race** (defect — cut evaluation time, ship the more capable model sooner to capture users, talent, and investment).

## 2. Write the payoff matrix (ordinal)

Rank each outcome 1st-best (4) to 4th-worst (1) from a single lab's private point of view:

- **(Race, Restrain) = T:** you ship first while the rival holds back. You capture the market, the headlines, the developer mindshare, and the next funding round. Best outcome. **T = 4**
- **(Restrain, Restrain) = R:** both hold back. The field moves at a safer pace, catastrophic-risk exposure is lower, and neither loses relative position. Second-best. **R = 3**
- **(Race, Race) = P:** both sprint. Evaluations get compressed, incident risk rises, margins get competed away in a compute arms race — but no one falls behind. Third. **P = 2**
- **(Restrain, Race) = S:** you hold back on principle while the rival ships. You lose users, talent, and capital, and the rival sets the norms anyway — so restraint bought you nothing and cost you the field. Worst. **S = 1**

```
                 Lab B: Restrain     Lab B: Race
Lab A: Restrain    R,R = 3,3          S,T = 1,4
Lab A: Race        T,S = 4,1          P,P = 2,2
```

## 3. Check whether it is a Prisoner's Dilemma

Ordering: **T (4) > R (3) > P (2) > S (1)** — the defining PD inequality holds. This is a true Prisoner's Dilemma, not Chicken: mutual racing (P) is the *second-worst* outcome, not the worst, because falling behind unilaterally (S) is worse than a shared sprint. (In Chicken, T > R > S > P, and mutual defection would be the disaster both most want to avoid — which is not how the labs actually rank being left behind.)

## 4. Ident

examples/flood-dresher-tucker-rand-1950.md

# Method in Action: Flood, Dresher, and Tucker — RAND, 1950

> *Example for the [prisoners-dilemma](../SKILL.md) skill.*

The Prisoner's Dilemma was not derived from a story. The story was attached to the matrix *after* the matrix had already been written down and observed to misbehave in a laboratory.

In **January 1950**, at the RAND Corporation in Santa Monica — then a Cold War strategic-studies institution — mathematicians **Merrill Flood** and **Melvin Dresher** were building game-theoretic tools to analyze nuclear stability. Their question: if Von Neumann and Morgenstern's *Theory of Games* (1944) predicted that rational players in zero-sum games would converge on saddle-point equilibria, what did rational play look like in **non-zero-sum** games — situations like arms races where both sides could win together or lose together?

Flood and Dresher wrote down a 2×2 non-zero-sum payoff matrix with a peculiar structure: each player had a strictly dominant strategy (each individually-rational move pointed the same way), but the strategy pair the rationality predicted produced an outcome that *both players preferred to avoid*. The matrix predicted self-sabotage by individually-rational actors.

To test whether real human reasoners would actually fall into this trap, Flood and Dresher ran an experiment. They recruited two colleagues: **Armen Alchian**, the economist (later UCLA), and **John D. Williams**, a RAND mathematician. The two subjects played the matrix 100 consecutive times, with each player privately recording their decision before each round. Flood preserved the full transcript, including the players' written commentary — a primary-source document of unusual richness for a 1950 social-science experiment. He published it later as *RAND Research Memorandum RM-789-1*, "Some Experimental Games" (1952; revised 1958).

The matrix Flood and Dresher used (in their published payoff units) was:

> "Player JW chooses row, Player AA chooses column... If both choose strategy 2 they receive (1/2, 1) respectively. If JW chooses 1 and AA chooses 2 they receive (-1, 2). If JW chooses 2 and AA chooses 1 they receive (0, 1/2). If both choose 1 they receive (1/2, 1)."

— Flood, M. M., "Some Experimental Games," RAND RM-789-1 (1952), p. 17. Reprinted in *Management Science* 5(1), pp. 5–26, October 1958. https://doi.org/10.1287/mnsc.5.1.5

The experimental results are the part the textbooks rarely emphasize. Over 100 rounds:

- **Alchian cooperated 68 times; Williams cooperated 78 times.**
- The Nash equilibrium prediction was that both would defect every round.

The subjects' written commentary, preserved verbatim in the RAND memorandum, captures the moment classical game theory hit its first empirical wall. Williams wrote during play: *"He's a shady character and doesn't realize we are playing a 3rd party, not each other."* Alchian wrote later: *"I'll be damned if I'll appease anybody."* What Flood and Dresher observed was that rational actors, playing the

Github ReposUpdated 2d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.