Persistent file-based planning for AI coding agents, surviving context loss and /clear.
planning-with-files is easy to set up with strong trust signals. Check agent compatibility and use-case fit before adding it to your workflow.
npx skills add OthmanAdi/planning-with-filesRun the command in your terminal.
Confirm that the skill files were added to your agent workspace.
Check the README requirements before invoking the skill in your agent.
Repository setup guidance
Strong trust signals; still review the README and permissions before production use.
Last commit was about 27 days ago.
26211 GitHub stars indicate community interest.
7 open issues signal maintenance load.
Codebase intelligence layer for AI coding agents: code health, git analytics, dead code detection, and architectural decisions via MCP.
Deep code indexing MCP server for AI agents with hybrid FTS5 + embedding search, call graphs, and multi-repo workspaces.
A project template that turns your repository into stable infrastructure for AI-assisted development.
This tool helps AI coding agents remember their plan even when their memory (context window) is cleared or crashes. It saves the plan, findings, and progress in three files on your computer and re-reads them every turn, so the agent can continue where it left off. It's like giving the agent a notebook that never gets lost.
planning-with-files is a skill for AI coding agents that implements persistent, file-based planning. It maintains three markdown files on disk: task_plan.md (the overall plan), findings.md (discoveries and notes), and progress.md (what has been done). These files are re-injected into the agent's context on every turn via a UserPromptSubmit hook, ensuring the plan survives context window resets, /clear commands, crashes, and context compaction. This Manus-style working memory approach prevents the agent from losing track of long-running tasks and reduces the need to re-explain goals. The skill includes an opt-in completion gate to ensure tasks are finished deterministically. It is designed to work across 60+ agents that support the Agent Skills standard, including Claude Code, Codex, Cursor, and OpenCode. The repository includes a benchmark showing a 96.7% assertion pass rate with the skill and a 3/3 win rate in blind A/B tests. Installation is available via npm, the Claude Code plugin marketplace, or npx skills. The project is MIT licensed and actively maintained.
Persistent file-based planning for AI coding agents, surviving context loss and /clear.
AI coding agent power users who frequently hit context limits., Developers building autonomous agents for complex, long-running tasks., Teams using Claude Code, Codex, Cursor, or other compatible agents in production.
The setup section provides repository-level starting guidance, not a guarantee of an independently verified installation. Check the official README and release notes.
The skill writes files to the local filesystem, which could be a security concern if used in untrusted environments. The plan files may contain sensitive information about the project, so ensure proper file permissions.
Similar or complementary options to evaluate include repowise, srclight, attractor-guided-engineering-template.
Aug 20, 2026
MIT license detected.
3 security/trust notes recorded.
Setup difficulty is 2/5.
26,211
Stars
2,197
Forks
7
Issues
MIT
License