Delivery Loop: Goal to Verified Close
Drive one deliverable from agreed goal and acceptance criteria to verified evidence and a real closeout, with nothing left orphaned.
Before & after
Same prompt, same agent — the only difference is the skill.
Sprints taking longer than expected is a common challenge, especially as teams scale. There are usually a few things going on — either estimates are too optimistic, scope creeps mid-sprint, blockers aren't being surfaced early, or the team is overloaded. To dig into what's happening in your case, I'd need to know more. How many sprints are we talking about? Is it consistent overrun, or just a recent thing? Do you track velocity metrics — points planned vs. completed? Are there particular types of work that tend to slip, or is it across the board? Also, how big is your team? Is this a new team, or an established one? Do you use Jira or another tracking tool where I could see the actual data? Once I understand the pattern, I can help you identify the root cause and suggest fixes. Is it estimation accuracy, planning discipline, technical debt slowing things down, or team capacity issues?
SPRINT HEALTH ANALYSIS & OPTIMIZATION PLAN CURRENT STATE (Last 8 sprints) Planned velocity: 120 points/sprint Actual completion: 78 points/sprint (65% delivery rate) Slippage: 42 points avg per sprint (scope creep) Burn-down curve: Days 1-5 on track, Days 6-10 plateau (daily standup shows low progress) VELOCITY TREND • Sprint 1: 110 pts (91%) • Sprint 2: 82 pts (68%) • Sprint 3: 72 pts (60%) • Sprint 4-8: avg 78 pts (65%) TREND: Declining velocity, hitting floor around 65-70% ROOT CAUSE ANALYSIS 1. ESTIMATION BIAS (40% of overrun) • 8-point stories avg 12-14 days (under-estimated by 40%) • Team members new; ramp-up costs not factored • Complex edge cases discovered mid-sprint 2. SCOPE CREEP (35% of overrun) • Unplanned bugs: 2-3 per sprint (avg 8 hours each) • Urgent asks from leadership mid-sprint: ~1 per sprint • Refactoring sneaks in (1-2 stories) 3. BLOCKERS (25% of overrun) • Waiting on design review: 2-3 days avg • API changes from backend team: 1-2 day delays • Merge conflicts / CI failures: 1 day overhead RECOMMENDED ACTIONS (Priority order) 1. Estimation calibration (implement immediately) • Use t-shirt sizing (S/M/L) + historical velocity chart • 5-point story = 5 ideal days + 1 day buffer for unknowns • Max story size: 8 points (anything bigger → spike first) • Retrospective: post-sprint story accuracy review 2. Scope gate (starts next sprint) • No mid-sprint requests without sprint closure review • Urgent asks: follow interrupt protocol (trade with planned work) • Separate bug backlog (reserve 10% capacity) 3. Dependency mapping (within 2 weeks) • Identify design, backend, infra dependencies • Prep those items one sprint ahead • Daily standup: surface blockers by noon EXPECTED OUTCOME Sprint velocity: 100+ points (83%+ delivery) within 3 sprints. First improvements visible in Sprint 1 with estimation fixes alone.
About this skill
Projects rarely fail at planning; they fail at the ends. A fuzzy goal at the start, then 'shipped, moving on' with no sign-off, no owner, and a feature flag and contractor access left behind. This skill runs one deliverable through a gated loop and will not call it done until it is verified and closed.
What it does
- Turns a request into an outcome goal, 3-7 checkable acceptance criteria, and explicit out-of-scope
- Plans verifiable milestones with the riskiest unknown first and verification time included
- Keeps one delivery page current instead of scattered status updates
- Verifies every criterion against evidence, and treats staging-only checks as not verified
- Runs a closeout: written acceptance, operating owner, temporary access and flags removed, follow-ups owned, metric check scheduled
What you get
A living delivery page with criteria, evidence, milestones, decisions, and a closeout checklist, plus a pointer to the right specialist skill when a request is really about sprints, JQL, or status reporting.
Best for
Project and delivery leads, engineering managers, and ops owners who need projects to actually finish.
SKILL.md preview
---
name: pm-skills
description: Run one piece of delivery work end to end as a single loop - clarify the goal and acceptance criteria, plan the smallest path, track execution in one delivery page, verify against the criteria with evidence, and close it properly (sign-off, handover, loose ends, lessons) - and route side requests like sprint analytics, JQL, or status reports to the right specialist skill. Use when the user says "take this from goal to done", "run this project for me", "how do we know this is actually done?", "close out this project", "we shipped but nobody signed off", "help me drive this deliverable", or "which PM skill do I need for this?".
---
… (sign up to view the full skill)More business & ops skills
View all Business & Ops skills →Confluence and Jira Template Builder
Build one Confluence or Jira template people actually fill in: real-example structure, placeholders, roll-up reports, required fields.
Confluence Space Architect and Cleanup
Make a Confluence space findable: shallow page tree, label taxonomy, navigation macros, and a CQL audit of stale and orphaned pages.
Jira JQL, Board and Automation Expert
Correct JQL on the first try, plus boards, dashboards, workflows, and automation rules that avoid Jira's silent traps.
Sprint History Analyzer
Find trends across 4+ sprints from your export: say/do, velocity stability, scope creep, carryover, retro follow-through, forecasts.