pj-flow

Organizes long-running projects into persistent memory files and timestamped thread folders across conversation sessions.

Updated May 20, 2026
One-click install
npx skills add https://github.com/TeXmeijin/agent-skills --skill pj-flow-texmeijin
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pj-flow
Source: https://github.com/TeXmeijin/agent-skills/tree/main/.apm/skills/pj-flow
Command: npx skills add https://github.com/TeXmeijin/agent-skills --skill pj-flow-texmeijin

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Long projects lose context when a chat thread ends, forcing you to re-explain goals, background, and progress in every new session. This Skill maintains a per-project long-term memory file plus timestamped thread folders (OBJECTIVE.md / OUTPUT.md) inside your repository, so any new session can resume work with full context. ## Core Features & Use Cases - Project lifecycle management: Initialize, resume, close, archive, and inventory-triage projects stored under .claude/my-projects/<slug>/ (or a custom root), with a fixed frontmatter schema tracking status, branch, PR, and issue URLs. - Thread-based work units: Split work into concrete (research/spec) and implementation threads using copyable templates, with pbcopy kickoff prompts for seamless handoff to the next session. - Retroactive adoption & import: Convert an in-progress conversation, existing PR, Issue, or branch into a fully structured project after the fact, and migrate legacy project-management formats. - Use Case: You are midway through a multi-week feature across server and client repos. Ask the agent to resume your project slug, and it reads the long-term memory and latest thread output, then proposes the next concrete steps without any re-explanation. ## Quick Start Ask the agent to start a new pj-flow project for your current task, or say "resume <slug>" to continue an existing project from its saved memory and latest thread.

Frequently Asked Questions about pj-flow

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

FAQPage Schema
How do I keep project context across multiple AI chat sessions?

Store a long-term memory file (CLAUDE.md) per project plus timestamped thread folders with OBJECTIVE.md and OUTPUT.md. When resuming, the agent reads the memory and latest thread output, then summarizes next steps without needing re-explanation.

How do I turn an existing pull request into a managed project?

Use the import-from-artifacts flow: it collects PR data via gh pr view, infers a slug, generates the long-term memory from the PR title and body, and creates a post-hoc implementation thread listing changed files and CI status. Open questions are flagged for your confirmation.

Does pj-flow work with Codex as well as Claude Code?

Yes. The default storage root is .claude/my-projects/, but the skill supports custom roots such as .agents/my-projects/ for Codex or multi-runtime setups. The user-specified location always takes priority.

Where are project files stored and are they committed to git?

Files live under .claude/my-projects/<slug>/ by default and are expected to be gitignored because they may contain private URLs and internal notes. The skill checks git check-ignore before writing and warns you if the directory is tracked.

What happens to old threads when project requirements change?

Old OUTPUT.md files are never deleted. Small changes are absorbed in the current thread, medium changes are noted in the next thread's OBJECTIVE, and large pivots spawn a reconcrete thread that explicitly names the outdated thread it overrides.