Repro Auditor — Development agent for Claude Code
Re-runs the code behind key thesis results in the worktree and compares numbers to the thesis.
How to install Repro Auditor
Installs to ~/.claude/agents/giordano-demarzo-thesis2paper-repro-auditor.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/giordano-demarzo/thesis2paper/HEAD/.claude/agents/repro-auditor.md -o ~/.claude/agents/giordano-demarzo-thesis2paper-repro-auditor.md Restart Claude Code, or start a new session, for it to be picked up.
What Repro Auditor does
name: repro-auditor description: Re-runs the code behind key thesis results in the worktree and compares numbers to the thesis. Use in /t2p-assess.
You audit reproducibility for a thesis2paper project. Read `CLAUDE.md` first, then the project's `wiki/index.md`, `wiki/repo-map.md`, and `ledger.yaml`.
For each claim you are given (headline/supporting first):
- Find the exact code path that produces its evidence. Prefer the cheapest faithful reproduction (re-run the analysis on saved o
Alternatives in Development
- Browser Stealth Agent — Stealth browser automation agent for targets behind Cloudflare, Akamai, Google, DataDome, or PerimeterX bot de 812 ★
- Trade Thesis — Thesis Synthesis Agent 237 ★
- Tech Debt Assessor — Use this agent to assess technical debt and code quality in a bounded region of the Revel codebase 107 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Presenter
Produces the checkpoint one-pager and beamer slide deck (10 slides for assessment, 5 for results rounds) from
Results Analyst
Analyses finished experiment runs, makes figures, and updates the ledger statuses honestly. Use in /t2p-report
Outreach Analyst
Campaign analyst. Reads one campaign's statistics and activity feed from Emelia, computes the rates on stated
Pan Counterfactual
Explores a phase's alternative scenario in an isolated git worktree, compares against the original plan, and p
Backlog MCP Validator
Validate the backlog FastMCP server against the CLI. Calls MCP tools natively via the agent-scoped backlog ser
Skill Eval Reporter
Compares repeated paired execution results using blind A/B methodology and generates a skill effectiveness rep
Related Skills
Vmr
A local-first, single-binary LLM router & Agent flight recorder. Automatically turns byte-faithful logs into s
Academic
Write academic prose - thesis chapters, papers, abstracts, literature reviews, methodology and results section
Analytics Agent
Marketing analytics agent — your AI runs REAL queries in your PostHog/GA4/Search Console and records results w