planning-with-files

Organizes complex multi-step tasks using persistent markdown planning files on disk.

Updated May 18, 2026
One-click install
npx skills add https://github.com/fts-pro/FTS-MMIS-AFRIKA --skill planning-with-files-fts-pro
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: planning-with-files
Source: https://github.com/fts-pro/FTS-MMIS-AFRIKA/tree/main/.agent/skills/planning-with-files
Command: npx skills add https://github.com/fts-pro/FTS-MMIS-AFRIKA --skill planning-with-files-fts-pro

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Long agent sessions lose track of goals because context windows are volatile and limited. This Skill implements Manus-style file-based planning, treating the filesystem as persistent external memory so multi-step projects survive context resets, compaction, and /clear. ## Core Features & Use Cases - Three-File Planning Pattern: Creates task_plan.md (phases and decisions), findings.md (research and discoveries), and progress.md (session log) in your project directory. - Session Recovery: A session-catchup script detects unsynced context from previous Claude Code, Codex, or OpenCode sessions after /clear, and a PreCompact hook flushes progress before compaction. - Parallel Plan Isolation: Supports multiple concurrent plans under .planning/<date>-<slug>/ with an active-plan pointer and PLAN_ID pinning per terminal. - Injection Defense: Wraps injected plan content in BEGIN/END delimiters and offers optional SHA-256 attestation that blocks injection if task_plan.md is tampered with. - Use Case: When asked to plan a multi-phase research or build project, the agent creates the three planning files first, re-reads the plan before decisions, logs every error, and resumes cleanly after context loss. ## Quick Start Ask the agent to plan out your multi-step project using planning files, and it will create task_plan.md, findings.md, and progress.md in your project root before starting work.

Frequently Asked Questions about planning-with-files

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I keep an AI agent on track during long multi-step tasks?▼

Use file-based planning: create task_plan.md with phases before starting, re-read it before major decisions, and log progress and errors to disk. This keeps goals in the model's attention window even after dozens of tool calls.

How do I recover agent context after /clear or context compaction?▼

Run the session-catchup.py script with your project path. It scans previous Claude Code, Codex, or OpenCode sessions for work done after the last planning file update, then recommends running git diff --stat and re-reading the planning files.

Can I run multiple planning sessions in the same repository?▼

Yes. Run init-session.sh with a project name to create an isolated plan under .planning/<date>-<slug>/. Switch between plans with set-active-plan.sh or pin a terminal with the PLAN_ID environment variable.

How does the skill prevent prompt injection through plan files?▼

Hook-injected plan content is wrapped in BEGIN/END delimiters and tagged as data only. Optionally, attest-plan.sh stores a SHA-256 hash of task_plan.md, and hooks block injection with a PLAN TAMPERED warning if the file diverges.

When should I not use file-based planning?▼

Skip it for simple questions, single-file edits, and quick lookups. The pattern is designed for tasks with three or more steps, research work, or anything spanning many tool calls where state persistence matters.