epictetus

What problem does it solve?

Epictetus addresses the need for objective, evidence-backed codebase evaluations and post-experiment forensic reviews so teams and multi-agent systems can quantify software quality, surface regressions, and attribute authorship reliably.

Core Features & Use Cases

• Structured Audits: Produces machine-readable JSON and human Markdown reports with a standardized rubric covering architecture, correctness, testing, security, and maintainability.
• Process Evidence Integration: Optionally ingests tmux session logs and Marcus API timelines to correlate runtime behavior, task IDs, and agent actions with code findings.
• Authorship & Contribution Analysis: Infers contributors from git history, performs reachability-based contribution attribution, and grades individual agents on spec adherence and net contribution.
• Runtime Verification: Attempts runtime smoke tests (start app, hit endpoints, run CLI help) and notes missing dependencies or over-mocked tests.
• Use Case: Post-experiment audits of Marcus multi-agent runs to score agent output, identify coordination failures, and generate both project-scoped fixes and global Marcus recommendations.

Quick Start

Audit the repository at /path/to/project and produce the standardized JSON and Markdown audit reports including inferred contributors, scores, and recommendations.