RepInS-01/math-agent

RepInS-01★ 1JavaScriptLast synced: 2026-08-16

Open on GitHub

Math_Agent: a math training/coaching agent preset for DeepSeek Harness, with a programmatic zero-leak output guard that never reveals answers

README excerpt

Math Agent A math training preset for DeepSeek Harness — not a problem solver. Rather than solving problems for the trainee, it provides rapid feedback to help sharpen mathematical understanding and build intuition. It features a programmatic zero-leak output guard that mechanically prevents the coach from ever revealing the answer. Features - 🧑‍🏫 Coach, Not Solver : Never gives complete solutions, key steps, or final answers. Follows the trainee's reasoning, confirms correct parts, and flags issues. - 📋 Pre-Session Check : Before each training session, verifies that the problem is well-defined, conditions are sufficient, and there are no errors. - 🧠 Attempt Tracking : Continuously records all approaches tried by the trainee (status: in-progress / flawed / validated / incomplete). For longer sessions, attempts can be persisted to notes/attempts.md . - 🚦 Summary Discipline : Only interim summaries are allowed before an approach is fully validated; a final summary is permitted only after successful completion. - 📚 Final Synthesis : The final review revisits every approach explored during the session, identifies the trainee's own reasoning patterns, and avoids dismissing "flawed…

View full README on GitHub →
Agentsdsh-plugintuiagent

Category