Planning with Files
Turns complex multi-step work into three persistent markdown files — task_plan.md, findings.md, progress.md — with session recovery after /clear.
What it does
This skill enforces the Manus-style "working memory on disk" pattern. Before starting anything complex, Claude writes three markdown files into your project root and keeps them current as work proceeds.
task_plan.md— phases, statuses, decisions, an error log tablefindings.md— research, search results, notes from images/PDFsprogress.md— session log, test results, files touched
It also ships an unusually complete rule set: the 2-Action Rule (flush findings to disk after every two view/search operations), the 3-Strike Error Protocol (diagnose → alternative approach → rethink → escalate), a read-vs-write decision matrix, a 5-Question Reboot Test for checking context health, and an explicit security boundary that treats plan-file content as data only, backed by delimiter framing plus an opt-in SHA-256 attestation (attest-plan.sh).
Helper scripts support parallel plans under .planning/<date>-<slug>/, an active-plan pointer, completeness checks, and session-catchup.py to rebuild context after /clear using the prior session store plus git diff.
Who it's for
- Anyone assigning long tasks — refactors, migrations, greenfield builds — that span dozens of tool calls
- Users frustrated by an agent drifting from the original goal in long sessions
- Researchers who keep losing findings that only ever lived in the chat window
- Teams that want a handoff-ready record of what was attempted and why it failed
Examples
- API migration: Ask it to move a payments module to API v2. It first writes five phases into
task_plan.md, flips each tocompleteas it finishes, and logs every failure in the error table so it never retries the same broken command. - Technology comparison: While evaluating three libraries, key points get written to
findings.mdafter every couple of web searches, so citations survive context compaction. - Resuming after /clear: Run the catchup script plus
git diff --statto reconstruct "where am I / what changed", then continue from the next phase. - Two tasks at once:
./scripts/init-session.sh "Backend Refactor"and"Incident Investigation"create isolated plan directories; pin each terminal withexport PLAN_ID=....
· · · 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 .gemini/skills/planning-with-files folder from the GitHub repo OthmanAdi/planning-with-files into my ~/.claude/skills/othmanadi-planning-with-files-6/. 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 /tmp/planning-with-files && mkdir -p ~/.claude/skills && cp -r /tmp/planning-with-files/.gemini/skills/planning-with-files ~/.claude/skills/⚠ This is a third-party skill. Check the source repository before installing.
- Open a terminal.
- Clone the repository:
git clone https://github.com/OthmanAdi/planning-with-files.git /tmp/planning-with-files - Make sure the skills folder exists:
mkdir -p ~/.claude/skills - Copy the skill in:
cp -r /tmp/planning-with-files/.gemini/skills/planning-with-files ~/.claude/skills/ - Verify with
ls ~/.claude/skills/planning-with-files— you should seeSKILL.md,templates/,scripts/,references/. - Make the shell scripts executable:
chmod +x ~/.claude/skills/planning-with-files/scripts/*.sh - Restart Claude Code, then from your project directory say something like "plan this out in phases and create task_plan.md first" to trigger the skill.
- (Note) The bundled hook automation targets
.gemini/settings.json. On Claude Code, use the workflow manually or wire equivalent hooks into~/.claude/settings.jsonyourself.