A Zettelkasten-based PKM system designed for software development teams and tech leads. Compatible with Claude Code and Cursor.
A set of agents, rules, templates, and a PKM skill that help you:
- Document decisions, devlogs, and learnings using atomic, connected notes
- Track sprints, OKRs, and weekly reviews with a lightweight goal cascade
- Use expert agents for Django, Next.js, DevOps, architecture audits, and more
- Keep your team's knowledge alive across sessions, tools, and people
No database. No app. Just markdown files, wiki-links, and AI agents that know how to use them.
# 1. Clone into your workspace
git clone https://github.com/balin-ar/dev-team-pkm.git /tmp/dev-team-pkm
# 2. Run setup (detects Claude Code / Cursor automatically)
bash /tmp/dev-team-pkm/scripts/setup.sh
# 3. Fill in your templates
# Edit USER.md, MEMORY.md, and goals/sprint.md with your team's infoThe setup script will:
- Copy agents to
.claude/agents/and/or.cursor/agents/ - Copy rules to
.claude/rules/and/or.cursor/rules/ - Install the PKM skill to
.agents/skills/pkm-documentation/ - Create
knowledge/andgoals/directories - Patch your
CLAUDE.mdwith PKM instructions (if not already there)
Every piece of knowledge is a single file with cross-references. Decision records, devlogs, learnings, project notes, conversation summaries — all connected via [[wiki-links]].
1:1 notes, sprint retros, incident reports, and team profiles. Everything a tech lead needs to keep context on their team.
OKRs → Sprint Goals → Weekly Reviews → Daily Notes. Top-down alignment, bottom-up tracking.
| Agent | What it does |
|---|---|
django-expert |
Django REST, Clean Architecture, Celery, PostgreSQL |
nextjs-expert |
Next.js 16, TypeScript, App Router, Playwright testing |
devops-expert |
Docker, CI/CD, DigitalOcean, Linux servers |
sprint-reviewer |
Generates sprint review reports from your dailies |
backlog-organizer |
Audits and prioritizes your backlog |
architecture-auditor |
Reviews decision records, flags tech debt |
vault-maintainer |
Keeps your knowledge vault clean and connected |
Ready-to-use templates for CLAUDE.md, MEMORY.md, USER.md, OKRs, sprints, and weekly reviews. Parameterized — fill in your team's details and go.
knowledge/
├── decisions/ # Technical & strategic decision records
├── devlog/ # What was built, fixed, shipped
├── learnings/ # Tools, techniques, patterns discovered
├── projects/ # Project notes (updated in-place)
├── conversations/ # Session summaries
├── daily/ # Daily index (links only)
├── team/ # 1:1 notes, retros, team profiles
└── goals/ # OKRs, sprints, weekly reviews
Document a decision:
"We decided to use Celery Beat instead of cron for scheduled tasks" → Agent creates
knowledge/decisions/2026-02-13-celery-beat-over-cron.mdwith context, options, reasoning, and links to the affected project.
Run a sprint review:
Use the
sprint-revieweragent on Friday → It reads your dailies, crosses with sprint goals, and generates a report: completed vs planned, velocity, blockers, carry-forward items.
Audit the vault:
Use
vault-maintainer→ Finds orphan notes, broken links, missing cross-references, and suggests connections.
- Setup Guide — Detailed installation instructions
- Customization — How to adapt to your team
- Workflow — Recommended daily/weekly workflow
Inspired by:
- obsidian-claude-pkm by ballred
- pkm-documentation skill from OpenClaw
MIT — see LICENSE.