Planning with Files
Manus-style persistent planning that keeps task_plan.md, findings.md and progress.md on disk so multi-step work survives /clear and auto-compaction.
What it does
This skill treats the context window as RAM and the filesystem as disk. For any task that needs more than a few tool calls, it creates and continuously maintains three markdown files in your project directory:
task_plan.md— phases, status, next step, decision journalfindings.md— research and discoveries (especially fragile multimodal content from images, PDFs, browsers)progress.md— session log, test results, errors encountered
It also registers hooks (UserPromptSubmit, PreToolUse, PostToolUse, Stop, PreCompact) that re-inject the plan each turn, and ships session-catchup.py to rebuild context after /clear. Extras include a 3-strike error protocol before escalating to you, isolated .planning/<date-slug>/ directories for parallel tasks, SHA-256 plan attestation, and opt-in autonomous / gated modes for long unattended runs.
Who it's for
- Anyone handing Claude long refactors, builds, or research spanning many tool calls
- Users frustrated by agents forgetting the goal or repeating the same failing action
- Developers running several parallel tasks in one repo from different terminals
- Teams that want a durable written record of what was tried and learned
Examples
- Large refactor: run
init-session.sh "Backend Refactor", work phase by phase; when auto-compaction hits, the agent re-reads the plan files and resumes exactly where it stopped. - Recovery after
/clear: invoke the skill and it compares the catch-up report againstgit diff --stat, then writes the missing work back into the planning files. - Long research task: the 2-action rule forces findings to disk after every two browse/view operations, so screenshot and PDF insights are never lost.
· · · Install guide · · ·
Install in the Claude app (no terminal)
- Download the ZIP with the button below.
- In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
- Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
Install in Claude Code
Let Claude do it — paste this into Claude Code
Install the skill I found on Claude Skill Mart. Copy the .pi/skills/planning-with-files folder from the GitHub repo OthmanAdi/planning-with-files into my ~/.claude/skills/pi-planning-with-files/. When it's done, tell me in one line what this skill can do.
Install with a command instead
git clone https://github.com/OthmanAdi/planning-with-files.git ~/.claude/skills/planning-with-files⚠ This is a third-party skill. Check the source repository before installing.
- Open a terminal and clone the skill:
git clone https://github.com/OthmanAdi/planning-with-files.git ~/.claude/skills/planning-with-files - If you also want the
/plan-goaland/plan-loopslash commands, install it as a plugin from inside Claude Code: run/plugin marketplace add OthmanAdi/planning-with-files, then/plugin install. - Verify the install with
ls ~/.claude/skills/planning-with-files— you should seeSKILL.md,scripts/andtemplates/. - Restart Claude Code so the hooks in the frontmatter are registered.
- Change into your project folder and ask something like "plan this out step by step"; the skill will create
task_plan.md,findings.mdandprogress.mdthere. - Optional: for parallel work run
sh ~/.claude/skills/planning-with-files/scripts/init-session.sh "Task name", and runscripts/plan-doctor.shif the hooks seem silent. - Note: planning files land in your project root, so consider adding them to
.gitignore, and start with the default (legacy) mode — gated mode can block session stop on supported hosts.