AGENTS.md

Guidance for coding agents working on Thinklogi Harness Evidence Collector in the thinklogi project.

Project overview

This guide covers harness/evidence.mjs. The evidence collector inspects app, logic, tests, package, build script, and AGENTS.md files to produce boolean feature flags consumed by the orchestrator.

Primary source files

• harness/evidence.mjs owns collectEvidence and readMaybe.
• harness/orchestrator.mjs consumes evidence flags to propose or complete tasks.
• public/app.js and public/decisionLogic.mjs contain the feature markers.
• tests/harness.test.mjs verifies the orchestrator behavior that depends on these flags.

Setup commands

npm run harness:init npm run harness:tick npm run harness:run npm test npm run build npm run dev

Change rules

• Keep evidence checks file-based and deterministic.
• Add new evidence flags only when the orchestrator and tests consume them.
• Do not treat missing files as fatal; readMaybe should return an empty string.
• Prefer checking stable implementation markers over fragile formatting.
• Keep evidence names positive and specific.

Testing

Use temporary project roots to cover missing files, present app shell, press-hold markers, create flow markers, reason chips, tests, build config, and Codex automation markers.

PR checklist

• collectEvidence returns every flag expected by runHarnessCycle.
• Missing files do not crash harness tick.
• npm test and npm run harness:tick pass.