trust-model-analyzer

A Claude Code skill that analyzes the trust model of a Solidity protocol and produces:

• trust-inventory.md, an internal privilege inventory (every privileged function mapped to its actor, with file:line), and
• tmaar.md, a client ready TMAAR (Trust Model and Assumed Risks) report.

It enumerates privileged roles, external trust dependencies, and upgradeability; rates each entity on a trust spectrum (Trustless, Limited trust, Trusted, Fully trusted); and, when deployment addresses are known, verifies holders on chain.

Install

ln -sfn "$(pwd)" "$HOME/.claude/skills/trust-model-analyzer"

Use

Inside a Claude Code session, invoke it directly with the slash command /trust-model-analyzer. Natural language also triggers it, for example "analyze the trust model of this repo" or "tmanalyze" (both are recognized trigger phrases in the skill description). No shell alias or CLI setup is needed.

Layout

• SKILL.md orchestrates six phases (scope, enumeration, per entity analysis, optional on chain verification, synthesis, output).
• detection-heuristics.md, trust-spectrum.md, onchain-verification.md, output-templates.md, context-template.md are the resource files.
• output-templates.md doubles as the format reference: skeletons with placeholder fields, no protocol content.

Outputs are written to the audited project's .context/outputs/N/.