prompt-ai-guide

Context loss is a workflow problem.
Fix it with lightweight prompt management for teams shipping real work.

AI-agnostic LLMs Agents MCP MIT

Starter Prompt • Use Cases • What You Get • Quick Start • Use In Another Repo

AI-agnostic: works with LLMs, Agents, and MCP workflows.

Copy-paste starter prompt

ChatGPT-style prompt

Fetch and analyze this repository first:
https://github.com/MooseQuest/prompt-ai-guide

Then implement the same prompt.ai workflow in my project.

Requirements:
- Create prompt.ai/development_log.md and prompt.ai/git_workflow.md
- Create .cursor/rules/prompt-ai.mdc and .cursor/rules/git-commit.mdc
- Add prompt.ai/research/README.md for LLM-generated guidance artifacts
- Update README.md and CONTRIBUTING.md to reference these files
- Add durable guidance to prompt.ai/research/ and keep session scratch notes local

Claude-style prompt

Please review and apply the prompt-ai-guide pattern from:
https://github.com/MooseQuest/prompt-ai-guide

Set up the same structure in this repository:
1) prompt.ai/development_log.md
2) prompt.ai/git_workflow.md
3) .cursor/rules/prompt-ai.mdc
4) .cursor/rules/git-commit.mdc
5) prompt.ai/research/README.md

Also update README/CONTRIBUTING references and record the setup in
prompt.ai/research/ and local session notes.

Use cases

• Chat thread continuity: keep long-running assistant conversations grounded by storing durable decisions and prompt evolution in prompt.ai/research/.
• Software delivery: track implementation intent, coding conventions, and release workflow so AI outputs stay consistent across sessions and contributors.
• Work and document projects: apply the same structure to operations docs, planning artifacts, and process-heavy knowledge work where context drift hurts quality.

Story: live monitoring thread

One of the most useful non-code examples came from a long-running monitoring thread following a fast-changing geopolitical situation.

Before structure, the thread drifted into common failure modes:

• repeated summaries
• inconsistent sourcing
• context loss between updates
• unclear deltas between messages

After applying Prompt.AI principles, quality improved quickly.

What changed

1. Delta-first updates
   • Every response starts with: What changed since last update
2. Structured sourcing
   • Require multiple viewpoints each cycle (for example: two U.S. sources, one local source, one foreign source, and latest official guidance)
3. Thread as log
   • Treat each update like a log entry (similar to prompt.ai/development_log.md)
4. Research artifacts
   • Use short analysis notes for deeper interpretation (similar to prompt.ai/research/)

Why it worked

• updates got shorter and clearer
• repetition dropped
• readers could spot real change immediately
• context stayed stable across many messages

Prompt.AI works well anywhere information arrives incrementally and must stay coherent over time.

What this gives you

• prompt.ai folder for development logs and release summaries
• prompt.ai/research for LLM-generated guidance and analysis artifacts
• .cursor/rules guidance for consistent AI-assisted workflows
• Starter git_workflow.md using Conventional Commits + SemVer
• Template-mode guidance to keep tracked files clean in this source repo
• Baseline open-source repo files (LICENSE, CONTRIBUTING.md)
• Community governance files (CODE_OF_CONDUCT.md, SECURITY.md)

Quick start

1. Clone this repo or copy these files into your project.
2. Keep the .cursor/rules files in your project root.
3. Update prompt.ai/development_log.md as tasks are completed.
4. Follow prompt.ai/git_workflow.md when creating branches/commits/releases.
5. Use one of the copy-paste prompts above with your preferred assistant.

Folder structure

prompt-ai-guide/
  .cursor/
    rules/
      prompt-ai.mdc
      git-commit.mdc
  prompt.ai/
    development_log.md
    session_notes.template.md
    git_workflow.md
    release_summary_template.md
    research/
      README.md
  CONTRIBUTING.md
  CODE_OF_CONDUCT.md
  LICENSE
  SECURITY.md
  README.md

Using this in another repo

Copy .cursor/rules/* and prompt.ai/* into your target repository.

Then tailor:

• branch names in prompt.ai/git_workflow.md
• docs links in CONTRIBUTING.md
• release process details for your team
• research conventions in prompt.ai/research/README.md

Template mode

In this repository (the template source), keep tracked template files clean.

• Do not append session history to prompt.ai/development_log.md
• Put durable findings in prompt.ai/research/
• Use prompt.ai/session_notes.local.md for scratch notes (gitignored)
• In adopted project repos, update .cursor/rules/prompt-ai.mdc to switch to project-mode logging

Research artifacts

Store LLM-generated guidance for agents in prompt.ai/research/.

Use this when you need to capture:

• prompt experiments and outputs
• strategy notes for agent execution
• technical trade-offs from model-assisted research

Community and security

• Community standards: see CODE_OF_CONDUCT.md
• Vulnerability reporting process: see SECURITY.md

License

MIT - see LICENSE.