plan-work

Decompose feature asks into GitHub milestones, epics, and estimated issues with projected timelines.

1|1|Updated Aug 2, 2026
One-click install
npx skills add https://github.com/blac9216/waypoint --skill plan-work-blac9216
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: plan-work
Source: https://github.com/blac9216/waypoint/tree/main/.claude/skills/plan-work
Command: npx skills add https://github.com/blac9216/waypoint --skill plan-work-blac9216

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? Turning a vague feature request into executable GitHub work items is slow and error-prone: issues lack provable acceptance criteria, estimates are guesses, and milestones have no defensible dates. This Skill converts an ask into a fully structured plan — research lanes, an owner decision record, domain epics, right-sized issues, and a data-calibrated roadmap projection — filed directly on GitHub. ## Core Features & Use Cases - Ask classification and research lanes: Classifies each request as spike, bounded, or architectural, and dispatches parallel research lanes whose findings are posted as issue comments and ratified by the owner before any design work begins. - Decomposition into provable issues: Files milestones, domain epics, and issues with acceptance criteria a reviewer can verify at merge, area labels, blocked-by dependencies, and per-issue size estimates, splitting any issue estimated L before filing. - History-calibrated estimation and timeline projection: Runs read-only scripts against repository history to build cycle-time calibration tables, then computes critical paths and proposes milestone due dates from observed parallelism. - Use Case: A product owner says "plan the offline bundle feature." The Skill researches open questions, interrogates the owner for decisions, files a milestone with epics and estimated issues in Backlog, and presents a projected end date with the sample sizes behind every number. ## Quick Start Ask the agent to plan a feature or decompose an epic, for example: "Plan the cross-enclave transfer feature and project when the milestone will land."

Frequently Asked Questions about plan-work

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

FAQPage Schema
How do I break a feature request into GitHub issues with estimates?▼

Invoke the plan-work skill with the feature ask. It classifies the request, optionally runs research lanes, interrogates the owner for decisions, then files a milestone, domain epics, and issues with acceptance criteria, area labels, blocked-by links, and per-issue size estimates in Backlog.

How to estimate issue cycle time from GitHub repository history?▼

Run scripts/history.sh, which is read-only and collects merged PRs, review rounds, labels, and cycle hours for closed issues over the last 90 days. It outputs a calibration table by area and size bucket with p50/p80 cycle hours, falling back to documented defaults when a bucket has fewer than three samples.

Does the planning skill commit files to the repository?▼

No. Research findings, decision records, and estimates live on GitHub issues, epics, and milestones as comments and body sections. The skill's rule is to commit nothing; only scratch files and GitHub objects are written.

Can I re-project milestone dates when priorities change?▼

Yes. Use the --resequence flag or run scripts/timeline.sh with --milestones or --milestone to recompute critical paths and proposed due dates for selected milestones only, holding everything else fixed. The agent applies due_on changes after owner confirmation.

What happens when an issue has a malformed estimate value?▼

timeline.sh treats an unparseable or over-ceiling est. cycle value like a missing estimate: it falls back to the issue's size default (or the M default with no size) and reports the issue number, offending text, and cause on stderr without aborting the run.

When should I not use this planning workflow?▼

Skip it for spike questions that only need an answer and for bounded asks that fit one or two issues — those use a short interrogation and direct filing. It is designed for repositories running the github-workflow process; other repositories get thinner calibration data.