agentleFS
joris887joris887/exosuit/.claude/skills/retrospective/SKILL.md

retrospective

Skill9 starsChanged 3 months agoMITView on GitHub

Sprint retrospective — consumes progress.md Metrics table, adds leading-lagging analysis, anti-pattern detection, and 4Ls framework.

73 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use dev-story, team-level and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • lilangMaxSkill

    dev-story

    Read a story file and implement it. Loads the full context (story, GDD requirement, ADR guidelines, control manifest), routes to the right programmer agent for the system and engine, implements the code and test, and confirms each acceptance criterion. The core implementation skill — run after /story-readiness, before /code-review and /story-done.

    Also in 12 projects that use this

  • CoralGameSkill

    team-level

    Orchestrate level design team: level-designer + narrative-director + world-builder + art-director + systems-designer + qa-tester for complete area/level creation.

    Also in 12 projects that use this

  • CoralGameSkill

    team-narrative

    Orchestrate the narrative team: coordinates narrative-director, writer, world-builder, and level-designer to create cohesive story content, world lore, and narrative-driven level design.

    Also in 12 projects that use this

  • lilangMaxSkill

    create-stories

    Break a single epic into implementable story files. Reads the epic, its GDD, governing ADRs, and control manifest. Each story embeds its GDD requirement TR-ID, ADR guidance, acceptance criteria, story type, and test evidence path. Run after /create-epics for each epic.

    Also in 11 projects that use this

  • +7 more from lilangMax
  • +1 more from CoralGame

The file

Show its rules and text

We split this skill into the rules it gives an agent

25 rules. 0 are common, 15 are distinctive, and 10 are very specific to this project.

  • 0

    Common

    Most of the 62 edited versions we read keep these.

  • 15

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run a sprint or weekly retrospective:

    line 16
  2. Distinctive:

    Read sprint specs and progress to build a complete picture:

    line 20
  3. Very specific:

    Read docs/progress.md → ## Sprint History table for trend data

    line 22
  4. Very specific:

    Read the sprint spec(s) being reviewed: docs/sprints/sprint-N.md — especially the ## Outcome and ## Decisions sections

    line 23
  5. Very specific:

    Read docs/progress.md → ## Metrics table and display it directly — sprint-end already computed these values.

    line 31
  6. Very specific:

    Read docs/progress.md → ## Sprint History table.

    line 41
  7. Very specific:

    Read references/metrics-analysis.md for the leading/lagging classification.

    line 60
  8. Very specific:

    Read references/metrics-analysis.md for the four anti-pattern definitions.

    line 70
  9. Distinctive:

    If feature stories are <50% of total, flag: "Less than half of sprint work on features — investigate maintenance burden." </IF>

    line 91
  10. Very specific:

    Run scripts/pm/metrics.sh --churn and display the high-churn files.

    line 97
  11. Distinctive:

    Run the metrics script for framework-specific data not captured in progress.md:

    line 102
  12. Very specific:

    Read sprint spec ## Outcome → Ground rules field across recent sprints:

    line 113
  13. Very specific:

    Read docs/technical-debt.md and recent sprint spec(s) ## Outcome → Debt delta fields:

    line 122
  14. Distinctive:

    Use the 4Ls format:

    line 144
  15. Distinctive:

    Consider: smooth processes, good decisions, effective patterns

    line 148
  16. Distinctive:

    Consider: technical discoveries, process improvements, gotchas found

    line 152
  17. Distinctive:

    Consider: tools, information, clarity, time

    line 156
  18. Distinctive:

    Consider: automation, better docs, different approach

    line 160
  19. Distinctive:

    Should the goal format change for next sprint?

    line 169
  20. Distinctive:

    Should the session capacity estimate change for next sprint?

    line 174
  21. Very specific:

    Rule updates needed: Should any patterns be added to CLAUDE.md, GROUND_RULES.md, or .claude/rules/ based on this sprint's experience?

    line 204
  22. Distinctive:

    If patterns or gotchas were discovered:

    line 219
  23. Distinctive:

    Add to CLAUDE.md for future sessions

    line 221
  24. Distinctive:

    Update any incorrect conventions

    line 222
  25. Distinctive:

    Add new commands if needed

    line 223

“Common” is measured: kept by at least 6 in 10 of the 62 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

retrospective

Run a sprint or weekly retrospective:

1. Gather Data

Read sprint specs and progress to build a complete picture:

  • Read docs/progress.md → ## Sprint History table for trend data
  • Read the sprint spec(s) being reviewed: docs/sprints/sprint-N.md — especially the ## Outcome and ## Decisions sections
  • Git log for commits and their messages
  • Any blockers or issues encountered (from sprint spec ## Notes and ## Decisions)

2. Metrics Dashboard

2.1. Progress Metrics (single source of truth)

Read docs/progress.md → ## Metrics table and display it directly — sprint-end already computed these values.

MetricCurrentTargetTrendStatus
[Copy all rows from progress.md Metrics table]

Sprint note: [Copy from progress.md if present]

2.2. Sprint-over-Sprint Comparison

Read docs/progress.md → ## Sprint History table. Extract the last 2 rows (current and previous sprint):

MetricThis SprintPreviousΔDirection
Goal achieved[✅/❌][✅/❌]—[streak]
Tasks completed[X][Y][+/-][↑/↓/→]
Cycle time[X.Xd][Y.Yd][+/-][↑/↓/→]
Change failure rate[X%][Y%][+/-][↑/↓/→]
Test coverage Δ[+X%][+Y%][+/-][↑/↓/→]
Code churn ratio[0.XX][0.YY][+/-][↑/↓/→]
AI effectiveness[0.XX][0.YY][+/-][↑/↓/→]
Sprint satisfaction[X/5][Y/5][+/-][↑/↓/→]

Also extract from sprint spec (docs/sprints/sprint-N.md → ## Outcome):

  • Sprint churn: % stories added/removed mid-sprint (target <20%, >40% = broken planning)
  • Done-to-commit ratio: completed/planned (80% healthy, >95% = under-committing, <65% = over-committing)

2.3. Leading vs Lagging Analysis

Read references/metrics-analysis.md for the leading/lagging classification.

From the Sprint History's last 3 rows, assess the trend direction for each group:

  • Leading indicators (churn, coverage Δ, satisfaction): [improving / stable / degrading]
  • Lagging indicators (CFR, tasks): [improving / stable / degrading]

Flag any divergence — see the reference file for interpretation guidance. This is the most actionable analysis: leading indicators predict what lagging will show in 1-2 sprints.

2.4. Anti-pattern Detection

Read references/metrics-analysis.md for the four anti-pattern definitions. Check each against Sprint History data:

  1. Perception gap: satisfaction ≥4 AND (CFR or churn rising 2+ sprints)?
  2. Task fragmentation: tasks rising AND cycle time shrinking 2+ sprints?
  3. Weak tests: coverage Δ positive AND CFR rising 2+ sprints?
  4. Maintenance spiral: <50% feature stories AND churn rising?

Flag any detected patterns with specific metric evidence.

2.5. Feature Time Ratio

<IF condition="docs/sessions/.activity-log.jsonl exists"> Parse `docs/sessions/.activity-log.jsonl` for story `status-change` events with `story_type` field. Count completed stories by type:
TypeCount%
feature[X][X%]
bugfix[Y][Y%]
refactor[Z][Z%]
other[W][W%]

If feature stories are <50% of total, flag: "Less than half of sprint work on features — investigate maintenance burden." </IF>

2.6. Hotspot Investigation (conditional)

<IF condition="Code churn ratio in Metrics table is 🟡 or 🔴"> Run `scripts/pm/metrics.sh --churn` and display the high-churn files. Cross-reference with sprint stories — are these files being actively developed, or are they unrelated files being destabilized? </IF>

2.7. Framework Metrics

Run the metrics script for framework-specific data not captured in progress.md:

bash scripts/pm/metrics.sh

Reports: skill success/failure rates, per-skill breakdown, tool usage, rule triggers. Identify: bottleneck skills (high failure), rework patterns (high Bash ratio), underutilized gates.

<IF condition="docs/reference/GROUND_RULES.md exists"> ### 2.8. Ground Rule Compliance

Read sprint spec ## Outcome → Ground rules field across recent sprints:

  • Violation trend — increasing violations signals architectural drift
  • Repeat violations — same rule violated multiple sprints: rule may be unrealistic or enforcement insufficient
  • Coverage gaps — rules never checked suggest enforcement isn't working</IF>
<IF condition="docs/technical-debt.md exists and has active or resolved items"> ### 2.9. Technical Debt Health

Read docs/technical-debt.md and recent sprint spec(s) ## Outcome → Debt delta fields:

  • Active inventory: [N] total — [N] critical / [N] high / [N] medium / [N] low
  • AI-origin ratio: [N] of [total] active items ([X%]) have origin: ai-generated
  • Growth trend: Sum debt deltas from sprint outcomes in the review period. Net debt growing, stable, or shrinking?
  • Resolution velocity: Items resolved since last retro. Average time from Since date to Resolved date.
  • Stale items: Any active items older than 90 days with no linked story? Flag as neglected.
  • Interest escalation: Any items whose interest changed from Stable → Growing since last review? These are compounding.
  • Estimate accuracy: For resolved items, compare Effort actual vs Effort estimate. Systematic bias?
<IF condition="Section 2.6 (Hotspot Investigation) produced results"> Cross-reference high-churn files from 2.6 with debt item Locations. Overlap means debt in actively changing code — it's compounding through every edit. </IF>

Flag actionable patterns:

  • Net debt growing 2+ sprints → "Debt accumulating faster than resolution — increase debt allocation in next sprint"
  • AI-origin >50% → "AI is primary debt source — review quality gates, coding standards enforcement, and AI-specific review depth"
  • Stale items >3 → "Debt register going stale — items not converting to sprint stories. Check sprint-start debt health check."</IF>

3. Retrospective Framework

Use the 4Ls format:

Liked (What went well?)

  • Consider: smooth processes, good decisions, effective patterns

Learned (What did we discover?)

  • Consider: technical discoveries, process improvements, gotchas found

Lacked (What was missing?)

  • Consider: tools, information, clarity, time

Longed For (What do we wish we had?)

  • Consider: automation, better docs, different approach

4. AI-Assisted Sprint Reflection

Sprint-goal-anchored questions (answer each with evidence, not impressions):

Sprint Goal Effectiveness

  • Did the sprint goal help scope decisions during story execution? Was it specific enough for AI to use as decision context?
  • Were there moments where the sprint goal was ignored or forgotten? What caused that?
  • Should the goal format change for next sprint? (e.g., too vague, too compound, too narrow)

Session & Capacity Accuracy

  • How many sessions were actually used vs estimated? Were S/M/L sizes accurate?
  • Which stories took more sessions than sized? Why? (complexity surprise, external dependency, scope change)
  • Should the session capacity estimate change for next sprint?

AI Execution Quality

  • Where did AI-generated code create review bottleneck? (time spent reviewing vs writing)
  • Were any tests weakened or deleted? Were quality gates effective at catching this?
  • Context management: Did we hit limits? Was the plan-then-execute pattern followed?
  • TDD discipline: Was test-first maintained, or did time pressure cause shortcuts?

Carry-Over & Churn Analysis

  • Were carried-over stories goal-critical or non-critical? (Non-critical carry-over = correct prioritization)
  • What was the sprint churn rate? If >20%, what caused mid-sprint additions?
  • Were any stories added mid-sprint that should have waited for backlog?

4.5. Architecture Decision Audit

Review whether architectural decisions made during this sprint are properly documented:

  • Scan commit messages and story plans for technology choices, pattern decisions, or trade-off discussions that aren't captured in docs/adr/
  • Check for "decisions in chat" — if the team discussed architectural choices during the sprint but didn't create ADRs, flag specific candidates (e.g., "Chose Redis over Memcached for caching — warrants an ADR")
  • Ground rule candidates — if a pattern was enforced informally during the sprint ("we always do X"), suggest formalizing it as a ground rule backed by an ADR

This prevents the "decisions in Slack" anti-pattern where architectural choices are made but never formalized.

4.6. AI Learnings

Review how AI tools performed during this sprint — what worked, what didn't, and what to update:

  • Effective patterns: Which prompting strategies, decomposition approaches, or AI workflows produced good results? Document for future sessions and teammates.
  • Anti-patterns observed: Did AI-generated code introduce specific defect types this sprint? (Check: phantom packages, code duplication, abstraction bypass, tautological tests, over-engineering — see docs/reference/TEAM_WORKFLOW.md → AI-Specific Review Checklist)
  • Comprehension check: Can every developer explain the AI-generated code they merged? If not, flag specific areas for code walkthroughs. (59% of developers ship AI code they don't fully understand — this is how comprehension debt accumulates.)
  • Rule updates needed: Should any patterns be added to CLAUDE.md, GROUND_RULES.md, or .claude/rules/ based on this sprint's experience?
  • Review effectiveness: Did the review layers (pre-commit → CI → AI review → human review) catch issues? Where did defects slip through?

Capture findings in the retrospective summary. Update CLAUDE.md with any new rules or anti-patterns discovered.

5. Action Items

Convert insights into specific actions:

ItemTypePriorityOwnerDue
[Action]Process/Tool/DocHigh/Med/LowSelf[Date]

6. Update CLAUDE.md

If patterns or gotchas were discovered:

  • Add to CLAUDE.md for future sessions
  • Update any incorrect conventions
  • Add new commands if needed

7. Retrospective Summary

## Retrospective - [Date]

### Period Covered
[Sprint X / Week of Y]

### Metrics Summary
[Copy the Metrics table from docs/progress.md — single source of truth]

### Key Signals
- Leading indicators: [improving/stable/degrading]
- Lagging indicators: [improving/stable/degrading]
- Anti-patterns detected: [list or "none"]
- Feature time ratio: [X]%

### Top 3 Positives
1. [Most impactful positive]
2.
3.

### Top 3 Improvements Needed
1. [Most important improvement]
2.
3.

### Action Items
[List from above]

### Process Adjustments
[Any changes to workflow]

Output the retrospective summary with action items.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.