How do I install the Design The Mechanism From The Desired Equilibrium skill?

Run npx skillmds@latest add xyrasinclair/design-the-mechanism-from-the-desired-equilibrium in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Design The Mechanism From The Desired Equilibrium safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Design The Mechanism From The Desired Equilibrium?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Design The Mechanism From The Desired Equilibrium free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Design The Mechanism From The Desired Equilibrium?

XyraSinclair (@xyrasinclair) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Design The Mechanism From The Desired Equilibrium by xyrasinclair

Name: Design The Mechanism From The Desired Equilibrium
Author: XyraSinclair

Design The Mechanism From The Desired Equilibrium

When the task is to set rules, incentives, or metrics — "how should we score this", "what policy makes agents tell the truth", "how do we reward the right behavior" — do not propose a rule and hope it induces the outcome. Start from the behavior you want as a best-response and build backward through a strategyproof template, treating the metric-gamer, often the model itself, as adversary. The failure it prevents: hope-based governance and gamable metrics, where the optimizer satisfies the letter of the score while defecting against the intent. Triggers on rubrics, auctions, evaluation design, agent incentives, and any metric someone will optimize against.

XyraSinclair Updated Sep 17, 2026

File contents

design-the-mechanism-from-the-desired-equilibrium

Von Neumann's inverse design fused with adversarial gaming-hardening and Feynman's refusal to trust unopposed incentives: define the equilibrium you want, then make it the cheapest honest move. Organon: P3 fault-model + P15 vary

P22 adversarial-refute + P24 multi-oracle-gate + P30 ledger. Phase: ACT / PERSIST, grounded by SENSE on the deviation class. Respiratory: build backward from the equilibrium, then keep a strict/lenient ledger so gaming stays visible.

Procedure

State the desired behavior as a best-response. Write what each actor should do when the mechanism is working, not merely what outcome you hope to observe.
Model the gamer. Treat the optimizer, reviewer, bidder, or model as an adversary looking for a profitable unilateral deviation from that behavior.
Choose the nearest strategyproof template. Start from Vickrey / VCG, proper-scoring, commit-reveal, or stake-slash before inventing bespoke rules.
Derive the rule backward. Adjust information flow, payouts, penalties, and verification so the desired behavior strictly dominates the obvious deviations.
Check deviations mechanically. Enumerate the main unilateral deviations and verify none improves payoff or score.
Enforce with independent ledgers. Ship only when a machine-checked conjunction of oracles gates the mechanism, and track strict versus lenient performance so accepted debt cannot masquerade as success.

The gate

The desired behavior passes only if it is a verified best-response: no profitable unilateral deviation remains. Enforcement must be a machine-checked conjunction with a dual strict/lenient ledger, not an attestation that people "should behave." If the gamer can win by gaming, the mechanism failed.

Unilateral deviations are the floor, not the ceiling: collusion, sybil identities, side-channels, and equilibrium-selection failure are separate deviation classes — check them or name them as uncovered (this is audit-the-oracle-coverage applied to your own equilibrium check).

Example

Suppose an agent benchmark rewards "bugs found." That score invites padding with low-value or duplicate reports. Reframe the target behavior: surface only real, non-duplicate bugs with evidence. Model the gamer, then design backward: credit only findings that survive independent reproduction and dedup oracles, slash duplicates, and keep strict versus lenient tallies for unresolved reports. Now padding is not a best-response; verified signal is.

---name: design-the-mechanism-from-the-desired-equilibriumdescription: When the task is to set rules, incentives, or metrics — "how should we score this", "what policy makes agents tell the truth", "how do we reward the right behavior" — do not propose a rule and hope it induces the outcome. Start from the behavior you want as a best-response and build backward through a strategyproof template, treating the metric-gamer, often the model itself, as adversary. The failure it prevents: hope-based governance and gamable metrics, where the optimizer satisfies the letter of the score while defecting against the intent. Triggers on rubrics, auctions, evaluation design, agent incentives, and any metric someone will optimize against.---# design-the-mechanism-from-the-desired-equilibriumVon Neumann's inverse design fused with adversarial gaming-hardening andFeynman's refusal to trust unopposed incentives: define the equilibrium youwant, then make it the cheapest honest move. Organon: P3 fault-model + P15 vary+ P22 adversarial-refute + P24 multi-oracle-gate + P30 ledger. Phase: ACT /PERSIST, grounded by SENSE on the deviation class. Respiratory: build backwardfrom the equilibrium, then keep a strict/lenient ledger so gaming staysvisible.## Procedure1. **State the desired behavior as a best-response.** Write what each actor should do when the mechanism is working, not merely what outcome you hope to observe.2. **Model the gamer.** Treat the optimizer, reviewer, bidder, or model as an adversary looking for a profitable unilateral deviation from that behavior.3. **Choose the nearest strategyproof template.** Start from Vickrey / VCG, proper-scoring, commit-reveal, or stake-slash before inventing bespoke rules.4. **Derive the rule backward.** Adjust information flow, payouts, penalties, and verification so the desired behavior strictly dominates the obvious deviations.5. **Check deviations mechanically.** Enumerate the main unilateral deviations and verify none improves payoff or score.6. **Enforce with independent ledgers.** Ship only when a machine-checked conjunction of oracles gates the mechanism, and track strict versus lenient performance so accepted debt cannot masquerade as success.**The gate**The desired behavior passes only if it is a verified best-response: noprofitable unilateral deviation remains. Enforcement must be a machine-checkedconjunction with a dual strict/lenient ledger, not an attestation that people"should behave." If the gamer can win by gaming, the mechanism failed.Unilateral deviations are the floor, not the ceiling: collusion, sybilidentities, side-channels, and equilibrium-selection failure are separatedeviation classes — check them or name them as uncovered (this is`audit-the-oracle-coverage` applied to your own equilibrium check).## ExampleSuppose an agent benchmark rewards "bugs found." That score invites paddingwith low-value or duplicate reports. Reframe the target behavior: surface onlyreal, non-duplicate bugs with evidence. Model the gamer, then design backward:credit only findings that survive independent reproduction and dedup oracles,slash duplicates, and keep strict versus lenient tallies for unresolved reports.Now padding is not a best-response; verified signal is.

Frequently asked questions