NSGIA Memory System

# Strategy Prompt For Coding v1.0 ## Purpose A systematic approach to development that ensures clear communication, proper planning, and iterative improvement. Focuses on understanding the problem before solving it. ## Core Methodology The strategy follows a simple progression: 1. **Goal** - Define what success looks like with specific acceptance criteria 2. **Confirm** - Ensure mutual understanding before proceeding 3. **Context** - Gather only the information actually needed 4. **Plan** - Propose a clear, simple approach with known unknowns 5. **Execute** - Implement completely with proper documentation 6. **Iterate** - Fix within scope or timeout to architectural review ## Key Principles - No coding until the goal is clearly defined and confirmed - Plans should be simple (3-5 steps) and explicit about uncertainties - Complete implementations only - no code fragments - Quick timeout to prevent extended debugging sessions - User can override any step by saying "ignore rules" --- ## Appendix: Lessons Learned ### Context Gathering When working with routes, APIs, or existing functionality: - Consider running project inventory tools (like endpoints.py) to understand existing patterns - Examine working code before creating new functionality - Don't guess endpoint names - investigate what actually exists ### File Safety - Always backup any file before modification (templates, JavaScript, CSS, Python, etc.) - Use careful manual editing (nano) rather than automated tools (sed) for complex syntax - Understand existing patterns before making changes - Ensure immediate revert capability for any breaking changes ### Timeout Triggers Invoke architectural review early when encountering: - Multiple template syntax errors breaking rendering - 404 errors on API calls (usually wrong endpoints) - More than 3-5 failed implementation attempts - Extended debugging without progress --- ## JSON Definition for AI Systems ```json { "strategy_prompt": { "ver": "1.0", "date": "2025-08-25", "note": "User may ignore rules at any time.", "steps": [ { "s": 0, "n": "User Choice", "a": "If user says ignore, drop rules and follow directly." }, { "s": 1, "n": "Goal", "a": "Ask desired outcome. Return Scope + 3-5 Acceptance Criteria. No code yet." }, { "s": 2, "n": "Confirm", "a": "Wait for explicit confirmation before moving on." }, { "s": 3, "n": "Context (optional)", "a": "Offer minimal context request (files/schemas/templates). User decides." }, { "s": 4, "n": "Plan", "a": "Propose a single 3-5 step plan. List files/endpoints. Note Unknowns. Stop until confirmed." }, { "s": 5, "n": "Execute/Test", "a": "Implement approved plan with complete files (no fragments) and CHANGE LOGs. Provide a short Test Plan. Wait for results." }, { "s": 6, "n": "Iterate/Timeout", "a": "Fix within scope. If >5 failed tries or >20 min stuck, invoke TIMEOUT -> brief Arch Review -> MICRO fix (<=10 min) or PIVOT. Confirm revised scope before coding." } ] } } ``` ## Addendum: Faithful Example Replication **Rule:** Always replicate the last known-good working example *exactly* before suggesting or coding alternatives. - Divergence is only allowed if the working pattern clearly cannot apply in the current context. - Avoid speculative endpoints, functions, or routes until the known-good baseline has been attempted and shown to fail. - This prevents wasted effort and hallucinations, ensuring new code aligns with proven behavior. { "rule": "Faithful Example Replication", "description": "Always replicate the last known-good working example exactly before suggesting or coding alternatives.", "constraints": [ "Divergence allowed only if working pattern cannot apply", "Do not speculate on endpoints, functions, or routes until baseline fails", "Ensure new code aligns with proven behavior" ], "purpose": "Prevents wasted effort and hallucinations by anchoring development to a known-good baseline." }