Skip to content

borghei/Claude-Skills / project-management/sprint-retrospective

sprint-retrospective

Verified

Data-driven sprint retrospectives from git history — velocity, cycle/lead time, contributor insights, and churn hotspots. Use to run a retrospective, analyze team velocity or throughput, or generate a retro report from commit history.

references/red-flags.md

View on GitHub
# Red Flags: Sprint Retrospective

> Common ways this skill's output goes wrong -- concrete examples, why they're bad, and how to fix them.

## How to use this document

Scan every retrospective format, output, and follow-up before treating it as the team's learning system. Each red flag has bad and good quoted examples.

---

## Red Flag 1: Retros That Surface No New Information

**Symptom.** Every retro produces the same list: "communication could be better, planning was tight, on-call was hard". No new insights.
**Why it's bad.** A retro is the team's learning ritual. When it surfaces nothing new, it has degenerated into a status meeting. The team's actual systemic issues stay invisible because the format is too shallow to catch them.
**Bad example:**
> "Retro 12, 13, 14: same three items in 'Did not go well' each time. Action: 'improve communication'. (No specifics.)"
**Good example:**
> "Retro 12: 'sprint goal slipped because acceptance criteria were vague on Story X.' Specific. Retro 13: 'engineering review of design specs happened too late; team got blocked.' Specific. Each retro produces a *new* root-cause hypothesis, tested in the following sprint."
**How to catch it.** Retro outputs identical to last 2 retros = change format (e.g. switch from Mad-Sad-Glad to 5-Whys, or do a deep dive on one ticket).

---

## Red Flag 2: Action Items Reborn Each Sprint

**Symptom.** Last sprint: "improve cross-team comms." This sprint: "improve cross-team comms." Sprint before that: "improve cross-team comms."
**Why it's bad.** Recurring action items signal that nothing actually changed. Either the action was never specific enough to be implementable, or it was assigned but not done, or it does not address the real problem. Either way, the team learns retros are toothless.
**Bad example:**
> "Sprint 12 action: 'improve cross-team communication.'
> Sprint 13 action: 'improve cross-team communication.'
> Sprint 14 action: 'really improve cross-team communication this time.'"
**Good example:**
> "Sprint 12: identified that the design-engineering handoff fails when there is no async-review window. Action: 'establish 48-hour async design review before sprint planning, owner = <name>, due = next sprint.' Sprint 13 retro: action verified; cycle-time-from-design-to-implementation dropped 30%. Next problem surfaced."
**How to catch it.** Action item language repeats verbatim across sprints = root-cause it, do not re-add.

---

## Red Flag 3: Retro Held Without the Team

**Symptom.** Scrum Master writes the retro alone using Jira data and sends out conclusions.
**Why it's bad.** Retros work because of the *conversation* -- the moment when one person's observation triggers another's "oh, I noticed that too" or "but actually, here is the data I have". An SM-authored retro loses this; the conclusions are one person's view dressed up as the team's.
**Bad example:**
> "Retro doc generated by SM from Jira + Slack analysis. Posted Friday. Team has 'opportunity to comment' but never does."
**Good example:**
> "Retro is a 60-min synchronous (or async-structured) discussion with the full team. Pre-work: each person reviews data + writes their own observations privately. During retro: share, cluster, vote, root-cause, decide. Outputs synthesized by SM; team owns them."
**How to catch it.** Retro doc with a single author and no team-contributed inputs = run a real retro.

---

## Red Flag 4: Blame Framing

**Symptom.** Retro output: "<name> did not finish the integration in time, which blocked the sprint."
**Why it's bad.** Blame retros destroy psychological safety. The next retro, no one will admit to anything. The team's learning system is broken.
**Bad example:**
> "Did not go well: '<name> dropped the ball on integration.' Action: '<name> takes ownership.'"
**Good example:**
> "Did not go well: 'integration slipped 1 week. Root cause analysis: the spec had two readings; the engineer assumed one, reviewer assumed another; the gap was discovered week 2. Systemic: we have no spec-review checkpoint between writing and starting work.' Action: 'add a 30-min spec-read-back before any integration story moves to In Progress.'"
**How to catch it.** Any retro output that names a person + a fault = rewrite to root cause + systemic fix.

---

## Red Flag 5: No Data Inputs

**Symptom.** Retro is "how do we feel?" with no metrics on the table.
**Why it's bad.** Feelings are a partial signal. Without data (cycle time, throughput, sprint goal completion, escaped bugs, on-call load), the team mistakes anecdotes for patterns and misses the systematic shifts.
**Bad example:**
> "Retro: 'I feel rushed.' 'I feel the work is fragmented.' 'I feel good.' (No data inputs.)"
**Good example:**
> "Retro opens with a 5-min data review: cycle-time trend (up 30% over last 3 sprints), sprint-goal completion rate (down from 85% to 60%), throughput stable, escaped-bug rate up. Discussion grounded: 'we are taking longer per item; why?'"
**How to catch it.** If the retro starts with feelings before data, run `velocity_analyzer.py` on the sprint's git history and `retro_report_generator.py` on its output to put a data slide on the table first.

---

## Red Flag 6: Mad-Sad-Glad Used For Every Single Retro

**Symptom.** Team has used the same Mad-Sad-Glad format for 18 consecutive retros.
**Why it's bad.** Any retro format gets stale. The same format reliably surfaces the same kinds of insights and misses the same blind spots. Rotating formats reveals different parts of the system.
**Bad example:**
> "Retro 1 through 18: Mad-Sad-Glad."
**Good example:**
> "Format rotation: Mad-Sad-Glad (sprint 1, 4, 7), Start-Stop-Continue (2, 5, 8), 4Ls (Liked / Learned / Lacked / Longed-for) (3, 6, 9), Sailboat (every 4th sprint), 5-Whys deep-dive (when a specific incident dominates), Fishbone (for systemic recurring issues). Format selected based on what data shows."
**How to catch it.** Same format > 4 retros in a row = rotate.

---

## Red Flag 7: Outputs Not Captured / Not Visible

**Symptom.** The retro happens, the team feels heard, no document is produced. A month later nobody remembers what was decided.
**Why it's bad.** Without a written output, the retro vanishes. Action items are not tracked. The next retro re-discovers old conclusions. The team's institutional memory degrades.
**Bad example:**
> "Retro held; no doc written; next retro held; nobody recalls last retro's outcomes."
**Good example:**
> "Every retro produces a 1-page doc with: data inputs, top 3 observations, decided actions (with owner + due date), parking-lot items, format used. Filed in `retros/sprint-NN.md`. Linked from the team's sprint dashboard. Next retro opens by reviewing last retro's actions."
**How to catch it.** No retro doc in the team's wiki / repo from the last sprint = capture going forward.

---

## Red Flag 8: All-Hands Retro Where Nobody Speaks

**Symptom.** Retro has 18 people; 4 talk; 14 are silent.
**Why it's bad.** A retro is only as good as the contributions it elicits. Silent participants signal psychological safety issues, or that the format is too group-think-inducing. The team gets a subset of the actual signal.
**Bad example:**
> "Retro: 18 attendees. Discussion dominated by 4 senior engineers. The 14 others contributed 'nothing to add'."
**Good example:**
> "Retro structure to surface quieter voices: (1) 5 min silent individual writing (everyone writes their own observations before discussion); (2) round-robin sharing (each person shares one item); (3) anonymous voting on what to dig into; (4) discussion on top items. For teams > 10, split into sub-groups for richer discussion."
**How to catch it.** Speaker share < 50% of attendees = restructure.

---

## Red Flag 9: Retros Skipped During Crunch

**Symptom.** Team is overloaded; cancels retros for 4 consecutive sprints. "We'll catch up later."
**Why it's bad.** Crunch is exactly when retros matter most -- they surface the over-commitment causing the crunch. Skipping them removes the team's last line of defense against the spiral.
**Bad example:**
> "Sprint 18, 19, 20, 21: retros canceled due to 'workload'. By sprint 22 the team is burned out and asks for a reorg."
**Good example:**
> "Retros are protected on the calendar. During crunch, retros are shorter (30 min) but never skipped. They explicitly surface 'how did we get here?' and feed scope-cut decisions."
**How to catch it.** > 1 skipped retro per quarter = treat skipping as a separate problem.

---

## Red Flag 10: Action Items Without Owners or Due Dates

**Symptom.** Actions: "Improve estimation." (No owner. No date.)
**Why it's bad.** Without owner + due date, the action is a wish, not a plan. By the next retro, nobody did it because nobody was responsible. The retro's credibility erodes.
**Bad example:**
> "Action: 'Make standups shorter.'"
**Good example:**
> "Action: 'Reduce standup to walk-the-board format starting Monday. Owner: <name> (scrum master). Due: end of next sprint. Definition of done: standup average duration < 12 min, measured over 5 standups.'"
**How to catch it.** Any retro action without Owner + Due Date + Definition of Done = re-write.

---

## Red Flag Quick Reference

| # | Anti-pattern | One-line check |
|---|---|---|
| 1 | Retros surface no new info | Change format if outputs repeat |
| 2 | Action items reborn | Root-cause, do not re-add |
| 3 | Retro held without the team | Real conversation, not SM monologue |
| 4 | Blame framing | Output names a person + fault = rewrite |
| 5 | No data inputs | Data slide opens every retro |
| 6 | One format used forever | Rotate formats every 4 retros |
| 7 | Outputs not captured | 1-page doc + tracked actions |
| 8 | Silent attendees | Speaker share >= 50% |
| 9 | Retros skipped during crunch | Treat skipping as a separate problem |
| 10 | Actions without owners | Owner + Due + DoD on every action |

## Related Reading

- `SKILL.md` -- the retrospective skill
- `scripts/velocity_analyzer.py --help` and `scripts/retro_report_generator.py --help` -- the sprint data and the retro report built from it
- `references/retro-formats.md` -- Mad-Sad-Glad, Start-Stop-Continue, 4Ls, Sailboat, Fishbone, 5-Whys
- Sibling skill: `scrum-master/` -- sprint analytics that feed the retro
- Sibling skill: `agile-coach/` -- broader agile-maturity coaching
- Sibling skill: `execution/quarterly-planning/` -- the quarterly retro at a longer cadence
- Sibling skill: `execution/summarize-meeting/` -- the structured output format
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 322kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 322kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 322kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 322kUpdated Claude CodeCodexVerifiedInstall