// the take
REAP is a prompt library for REA ("Reverse Engineer Anything"), a tool that lets an AI coding agent inspect binaries, firmware, APKs, Electron apps, games, kernels and websites. It's for developers already using REA who want ready-made prompts instead of writing their own, spanning everything from a basic feature trace to anti-cheat, DRM and vulnerability research.
Prompts consistently demand file/function/address evidence and force the agent to separate confirmed facts from hypotheses, which is a real, specific defense against LLM confabulation in reverse-engineering work rather than a vague quality gesture. The use-case sections escalate in a usable order (trace one feature → full clone spec → WASM port plan), so you can chain prompts into an actual workflow instead of picking one in isolation. The 'one feature per prompt, static analysis first' guidance reads like it came from actually running these against agents and hitting shallow, hand-wavy answers from broader asks.
There's no code to evaluate here, just prompt text — its entire value depends on REA, a third-party product this repo doesn't control or version against. Several sections (anti-cheat, DMA cheat research, cracking, DRM) carry real misuse potential, and the actual boundary against misuse is a single disclaimer paragraph plus the user's judgment, not anything enforced. Prompts reference specific REA workflows (verify_reconstruction, compare_application_versions, audit_residual_unknowns) with no changelog tying the repo to REA's release cycle, so they'll silently break if REA renames or drops them. Outside the one CSV walkthrough in the getting-started section, there's no example output showing any of these prompts actually produce the evidence-backed results they promise.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →