conduct-retrospective
À propos
Cette Compétence Claude facilite les rétrospectives structurées de projet ou de sprint en analysant les rapports d'état et les métriques pour identifier les succès et les axes d'amélioration. Elle génère des points d'action avec des responsables attribués et des échéances. Les développeurs doivent l'utiliser après les sprints, les jalons ou les événements significatifs pour systématiquement capitaliser sur les enseignements tirés.
Installation rapide
Claude Code
Recommandénpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/conduct-retrospectiveCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
Conduct a Retrospective
Facilitate a structured retrospective that reviews recent project execution, identifies what worked and what didn't, and produces actionable improvement items that feed back into project processes. This skill transforms raw project data into evidence-backed learnings with specific actions, owners, and due dates.
When to Use
- End of a sprint (sprint retrospective)
- End of a project phase or milestone
- After a significant incident, failure, or success
- Quarterly review of ongoing project processes
- Before starting a similar project (lessons learned review)
Inputs
- Required: Period under review (sprint number, date range, or milestone)
- Optional: Status reports from the review period
- Optional: Sprint velocity and completion data
- Optional: Previous retrospective actions (to check closure)
- Optional: Team feedback or survey results
Procedure
Step 1: Gather Retrospective Data
Read available artifacts from the review period:
- STATUS-REPORT-*.md files for the period
- SPRINT-PLAN.md for planned vs actual
- BACKLOG.md for item flow and cycle times
- Previous RETRO-*.md for open action items
Extract key facts:
- Items planned vs completed
- Velocity trend
- Blockers encountered and resolution time
- Unplanned work that entered the sprint
- Open action items from previous retrospectives
Got: Data summary with quantitative metrics (velocity, completion %, blocker count).
If fail: If no artifacts exist, base the retrospective on qualitative observations.
Step 2: Structure "What Went Well"
List 3-5 things that worked well, with evidence:
## What Went Well
| # | Observation | Evidence |
|---|------------|---------|
| 1 | [Specific positive observation] | [Metric, example, or artifact reference] |
| 2 | [Specific positive observation] | [Metric, example, or artifact reference] |
| 3 | [Specific positive observation] | [Metric, example, or artifact reference] |
Focus on practices to continue, not only outcomes. "Daily standups kept blockers visible" is more actionable than "We delivered on time."
Got: 3-5 evidence-backed positive observations.
If fail: If nothing went well, look harder — even small wins matter. At minimum, the team completed the period.
Step 3: Structure "What Needs Improvement"
List 3-5 things that need improvement, with evidence:
## What Needs Improvement
| # | Observation | Evidence | Impact |
|---|------------|---------|--------|
| 1 | [Specific issue] | [Metric, example, or incident] | [Effect on delivery] |
| 2 | [Specific issue] | [Metric, example, or incident] | [Effect on delivery] |
| 3 | [Specific issue] | [Metric, example, or incident] | [Effect on delivery] |
Be specific and factual. "Estimation was off" is vague. "3 of 5 items exceeded estimates by >50%, adding 8 unplanned days" is actionable.
Got: 3-5 evidence-backed improvement areas with stated impact.
If fail: If the team claims everything is fine, compare planned vs actual metrics — gaps reveal issues.
Step 4: Generate Improvement Actions
For each improvement area, create an actionable item:
## Improvement Actions
| ID | Action | Owner | Due Date | Success Criteria | Source |
|----|--------|-------|----------|-----------------|--------|
| A-001 | [Specific action] | [Name] | [Date] | [How to verify success] | Improvement #1 |
| A-002 | [Specific action] | [Name] | [Date] | [How to verify success] | Improvement #2 |
| A-003 | [Specific action] | [Name] | [Date] | [How to verify success] | Improvement #3 |
Each action must be:
- Specific (not "improve estimation" but "add estimation review step to grooming")
- Owned (one person accountable)
- Time-bound (due date within next 1-2 sprints)
- Verifiable (success criteria defined)
Got: 2-4 improvement actions with owners and due dates.
If fail: If actions are too vague, apply the "how would you verify this was done?" test.
Step 5: Review Previous Actions and Write Report
Check previous retrospective actions for closure:
## Previous Action Review
| ID | Action | Owner | Status | Notes |
|----|--------|-------|--------|-------|
| A-prev-001 | [Action from last retro] | [Name] | Closed / Open / Recurring | [Outcome] |
| A-prev-002 | [Action from last retro] | [Name] | Closed / Open / Recurring | [Outcome] |
Flag recurring items (same issue appearing in 3+ retrospectives) — these need escalation or a different approach.
Write the complete retrospective:
# Retrospective: [Sprint N / Phase Name / Date Range]
## Date: [YYYY-MM-DD]
## Document ID: RETRO-[PROJECT]-[YYYY-MM-DD]
### Period Summary
- **Period**: [Sprint N / dates]
- **Planned**: [N items / N points]
- **Completed**: [N items / N points]
- **Velocity**: [N] (previous: [N])
- **Unplanned Work**: [N items]
### What Went Well
[From Step 2]
### What Needs Improvement
[From Step 3]
### Improvement Actions
[From Step 4]
### Previous Action Review
[From Step 5]
---
*Retrospective facilitated by: [Name/Agent]*
Save as RETRO-[YYYY-MM-DD].md.
Got: Complete retrospective document saved with actions, evidence, and previous action review.
If fail: If the retrospective has no improvement actions, it's not driving change — revisit Step 3.
Validation
- Retrospective file created with date-stamped filename
- Period summary includes quantitative metrics
- "What Went Well" has 3-5 evidence-backed items
- "What Needs Improvement" has 3-5 evidence-backed items
- Improvement actions have owners, due dates, and success criteria
- Previous retrospective actions reviewed for closure
- Recurring issues flagged
Pitfalls
- Blame game: Retrospectives review processes and practices, not people. Frame issues as systemic, not personal.
- Actions without follow-through: The biggest retrospective failure. Always review previous actions before creating new ones.
- Too many actions: 2-4 focused actions are better than 10 vague ones. The team can only absorb so many changes.
- No evidence: "We feel estimation is bad" is opinion. "3 of 5 items exceeded estimates by 50%" is data. Always attach evidence.
- Skipping the positives: Only discussing problems is demoralizing. Celebrating wins reinforces good practices.
Related Skills
generate-status-report— status reports provide the data for retrospectivesmanage-backlog— improvement actions feed back into the backlogplan-sprint— retrospective learnings improve sprint planning accuracydraft-project-charter— review charter assumptions and risk accuracycreate-work-breakdown-structure— review estimation accuracy against WBS
Dépôt GitHub
Compétences associées
content-collections
MétaCette compétence propose une configuration éprouvée en production pour Content Collections, un outil axé sur TypeScript qui transforme des fichiers Markdown/MDX en collections de données typées de manière sûre avec une validation Zod. Utilisez-la lors de la création de blogs, de sites de documentation ou d'applications Vite + React riches en contenu pour garantir la sécurité de typage et la validation automatique du contenu. Elle couvre tout, de la configuration du plugin Vite et de la compilation MDX à l'optimisation des déploiements et la validation des schémas.
polymarket
MétaCette compétence permet aux développeurs de créer des applications avec la plateforme de marchés prédictifs Polymarket, incluant l'intégration d'API pour le trading et les données de marché. Elle fournit également une diffusion de données en temps réel via WebSocket pour surveiller les transactions en direct et l'activité du marché. Utilisez-la pour mettre en œuvre des stratégies de trading ou pour créer des outils traitant les mises à jour de marché en direct.
creating-opencode-plugins
MétaCette compétence aide les développeurs à créer des plugins OpenCode qui s'interconnectent avec plus de 25 types d'événements tels que les commandes, les fichiers et les opérations LSP. Elle fournit la structure du plugin, les spécifications de l'API événementielle et les modèles d'implémentation pour les modules JavaScript/TypeScript. Utilisez-la lorsque vous avez besoin d'intercepter, de surveiller ou d'étendre le cycle de vie de l'assistant IA OpenCode avec une logique personnalisée pilotée par les événements.
sglang
MétaSGLang est un framework de service LLM haute performance spécialisé dans la génération rapide et structurée pour les workflows JSON, regex et agentiques grâce à son cache de préfixe RadixAttention. Il offre une inférence nettement plus rapide, particulièrement pour les tâches avec des préfixes répétés, ce qui le rend idéal pour les sorties complexes et structurées ainsi que les conversations multi-tours. Choisissez SGLang plutôt que des alternatives comme vLLM lorsque vous avez besoin d'un décodage contraint ou que vous construisez des applications avec un partage étendu de préfixes.
