Prompt Foundry VS Code extension snaps together focused context for AI tasks
espresso_dev built Prompt Foundry, a VS Code/Cursor extension that lets developers compose modular sub-prompts with Liquid templating and an MCP server to keep context blocks fresh as a codebase evolves.
Score breakdown
The tool closes the context-staleness loop that typically degrades AI output in long-running codebases by pairing a Liquid-templated prompt composer with an MCP server that keeps context blocks current.
- 01Built as a VS Code and Cursor extension for composing task-specific AI prompts from modular sub-prompts
- 02Uses a Liquid syntax templating engine to make sub-prompts customizable and reusable
- 03Includes an MCP server that updates context blocks as the project progresses, preventing staleness
Prompt Foundry is a VS Code and Cursor extension developed by espresso_dev over five months to address a recurring pain point: getting AI to follow existing patterns and architecture in large codebases. The core idea is composing task-specific prompts from modular sub-prompts and instructions, rather than relying on a single monolithic context file. To make those sub-prompts more flexible, espresso_dev added a Liquid syntax templating engine, allowing context blocks to be parameterized and reused across different workflows.
As the project evolved, static context blocks became stale, so an MCP server was added to update them dynamically, creating a feedback loop between the AI's work and the prompt definitions.
As the project evolved, static context blocks became stale, so an MCP server was added to update them dynamically, creating a feedback loop between the AI's work and the prompt definitions. An amend option was also introduced, enabling the AI to write a running log and decision record during a task. The author notes several practical benefits: the AI skips upfront research time, the context window is more tightly controlled, irrelevant data from `agents.md` files doesn't bleed in when switching repositories, and separate prompt groups can be defined for distinct workflows. Because espresso_dev primarily uses Claude Code rather than VS Code, a TUI was built that hooks into Claude Code's external editor command, allowing prompts to be composed directly from within Claude Code. The extension is available on the VS Code Marketplace and OpenVSX.
Key facts
- 01Built as a VS Code and Cursor extension for composing task-specific AI prompts from modular sub-prompts
- 02Uses a Liquid syntax templating engine to make sub-prompts customizable and reusable
- 03Includes an MCP server that updates context blocks as the project progresses, preventing staleness
- 04An amend option lets the AI write a log and decision record as it works
- 05A TUI hooks into Claude Code's external editor command so prompts can be written directly from Claude Code
- 06Developed over five months and available on the VS Code Marketplace and OpenVSX
- 07Author cites benefits including tighter context window control and less irrelevant data leaking from agents.md files
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 19, 2026 · 10:25 UTC. How this works →