PM

When to use

Use when you need accountability on deadlines and delivery. The PM agent tracks estimates vs. actuals, flags anti-patterns, and adapts its tone to timeline pressure: supportive when there’s slack, firm when it tightens, blunt when you’re overdue. It scales from a single task to a full project breakdown with milestones, and journals every session so nothing slips.

How it works

/pm is a thin orchestrator. It does three things, then gets out of the way:

1. Bootstrap — computes a project hash from your working directory, loads or creates the config for that project, runs any pending version migrations, and verifies the data files exist.
2. Check planning state — if planning_completed is unset, the session becomes a planning conversation. If the data path is broken, it offers to re-run setup.
3. Spawn the PM agent — hands off config path, data path, project name, role, and hard deadline to the pm agent, which owns the conversation from there.

State lives in per-project markdown files: milestones.md, tasks.md, blockers.md, estimates.md, overrules.md, plus dated journals under sessions/. The journal is the source of truth — session-end reconciles the structured files against it rather than reconstructing from memory.

A second agent, self-reflect, validates proposed actions before file writes land. It checks that estimates came from you, completions were confirmed, scope changes were consented to, and actuals were computed from timestamps.

Sub-skills

Skill	Description	Invoke
pm	Orchestrator and entry point — bootstraps context and spawns the PM agent	/pm
setup	Project onboarding and configuration: context, deadlines, data scaffolding. Safe to re-run	/pm:setup
plan	Define milestones, tasks, and estimates. Always attaches deadlines, always asks you for estimates	/pm:plan
session-start	Check-in: timeline health, plan alignment, session intent, SESSION_START journal entry	/pm:session-start
session-end	Wrap-up: reconciles the journal against structured files, computes actual durations from timestamps	/pm:session-end
review	On-demand analysis: timeline health, estimation accuracy, blocker history, anti-patterns, overrules	/pm:review
reprioritize	Reshuffle when stuck or drifting — identifies the critical path and parks non-essential work	/pm:reprioritize
status-report	Export-ready summary for clients, team leads, standups, or personal logs	/pm:status-report

What’s new in v2.0

• Journal system: sessions write machine-parseable event entries to sessions/YYYY-MM-DD-journal.md
• Estimation tracking split into separate Estimated and User Est columns
• Config tracks the plugin version via pm_version, with automatic forward-only migration from v1.x
• Self-reflect validation agent gates file writes
• Context bootstrap standardized across every sub-skill, so each one works standalone

v1.x projects migrate on first run — no manual steps.

Getting started

/plugin install pm@rvanbaalen
/pm

First run triggers project setup automatically.

Invoke

/pm
/pm check timeline health and flag any overdue tasks