A set of agent skills that let an orchestrating AI agent delegate coding tasks to installed coding CLIs, then review the diff and commit it yourself.
OthmanAdi/planning-with-files
A file-based planning skill for AI coding agents that keeps task plans on disk so they survive context resets, /clear, and crashes.
- Stars
- 27,923
- Forks
- 2,313
- Open issues
- 11
- Last commit
- 3 days ago
What is it?
Planning with Files is a skill for AI coding agents that stores a task's plan, research notes, and progress in three markdown files inside the project: task_plan.md, findings.md, and progress.md. Because the state lives on disk rather than in the context window, an agent can resume at the current phase after /clear, context compaction, or a crash.
Lifecycle hooks re-inject the active plan context at the start of each turn, and a completion check reports whether the plan's phases are finished. Parallel tasks can use separate directories under .planning/, selected through an active-plan pointer. Automatic recovery reads only project files; reading local agent session records requires an explicit catchup mode.
The skill installs through npx skills, the Claude Code plugin marketplace, or npm, and has native plugin integrations for Claude Code, Codex CLI, Cursor, OpenCode, Hermes Agent and others, with translated variants in several languages.
Key features
- 01Stores plan, findings and progress as markdown files in the project
- 02Re-injects the active plan into each turn via lifecycle hooks
- 03Recovers the plan after /clear, compaction or session crashes
- 04Completion check reports whether all plan phases are finished
- 05Isolated plan directories for parallel tasks
- 06Installs across many coding agents via the Agent Skills standard
Use cases
- Long-running refactors or migrations that span many sessions
- Multi-step coding tasks where an agent must not lose track after /clear
- Keeping an auditable record of decisions and test results during agent work
In collections
Topics
FAQ
- What is planning-with-files?
- A file-based planning skill for AI coding agents that keeps task plans on disk so they survive context resets, /clear, and crashes. Planning with Files is a skill for AI coding agents that stores a task's plan, research notes, and progress in three markdown files inside the project: task_plan.md, findings.md, and progress.md. Because the state lives on disk rather than in the context window, an agent can resume at the current phase after /clear, context compaction, or a crash.
- Is planning-with-files free to use? What license does it have?
- planning-with-files is open source under the MIT license. Check the license terms before using it in commercial products.
- Is planning-with-files still maintained?
- Actively maintained. The last commit to planning-with-files was on Nov 4, 2026.
- What can I use planning-with-files for?
- Long-running refactors or migrations that span many sessions Multi-step coding tasks where an agent must not lose track after /clear Keeping an auditable record of decisions and test results during agent work
- Where can I get planning-with-files?
- The source code and installation instructions are on GitHub: https://github.com/OthmanAdi/planning-with-files
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Local-first persistent memory for AI coding agents, storing decisions and lessons in SQLite and serving them over MCP to Claude Code, Cursor and other clients.
Plain-language agent skills for Claude Code, Codex CLI, Cursor CLI and Gemini CLI that turn session reports into a one-screen verdict anyone can follow.
A curated directory of 1,000+ Agent Skills from official vendor teams and the community, for Claude Code, Codex, Gemini CLI, Cursor and more.
A spec-driven development workflow for AI coding assistants that agree on proposals, specs and tasks before any code is written.
A port of Poteto's pstack Cursor skill stack, giving Claude Code, Codex, Copilot, Pi and other agent harnesses structured workflows for coding tasks.