CHAIN PROMPT
# THIS IS A TEMPLATE — NOT AN ACTIVE PROJECT # Copy to your experiment directory and fill in sections. # Active experiments: chp-test-run/ and experiments/ # All experiments: see EXPERIMENT_INDEX.json # ──────────────────────────────────────────────────────── # [PROJECT NAME] — Chain Prompt Master File # This is the LIVING MASTER DOCUMENT for your project. # Pass this file in full at the start of every agent session. # # It contains: # - All confirmed design decisions (authoritative) # - Scientific constraints and citations # - Architecture rules # - File inventory # # If this file and any other file conflict, THIS FILE WINS. # Update this file whenever a design decision is locked in. ================================================================================ PROJECT IDENTITY ================================================================================ Name: [YOUR PROJECT NAME] Location: [PROJECT ROOT] Language: Python 3.11+ Status: [CURRENT PHASE] Purpose: [1-3 sentences describing the scientific question this project investigates] ================================================================================ CONFIRMED DESIGN DECISIONS ================================================================================ [Add decisions here as they are made. Each should have:] - What was decided - Why (citation or reasoning) - Date confirmed Example: DD01 — Agent update rule: synchronous (Nowak & May 1992) DD02 — Grid topology: 8-cell Moore neighborhood DD03 — Payoff matrix: b=1.8, c=1.0 (standard spatial PD) ================================================================================ ARCHITECTURE RULES ================================================================================ - Pure library: simulation has NO UI, NO print statements. All IO at the edges. - All randomness via seeded numpy.random.Generator. Same seed = identical results. - Models know nothing about engines — no circular imports. - Events are dicts: {type, step, description, outcome} - Structured logging via logging.getLogger(__name__). ================================================================================ FROZEN CODE ================================================================================ The following paths are FROZEN. Do NOT modify any file in these paths. New code must compose WITH frozen code, not modify it. frozen/ If unsure whether a file is safe to touch — do not touch it. ================================================================================ FILE INVENTORY ================================================================================ [List all project files with line counts and descriptions] | File | Lines | Description | |------|-------|-------------| | [path] | [N] | [what it does] | ================================================================================ CHANGE LOG ================================================================================ [Date] — [What changed and why]
fill the variables
This prompt has 1 variable. Pro fills them into a ready-to-paste prompt for you — no manual find-and-replace.
{type, step, description, outcome}
Unlock with Pro →when to use it
Community prompt sourced from the open-source GitHub repo kepiCHelaSHen/context-hacking (NOASSERTION). A "CHAIN PROMPT" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.
tags
roleplaycommunitygeneral
source
kepiCHelaSHen/context-hacking · NOASSERTION