Help us improve
Share bugs, ideas, or general feedback.
From tap-skills
Just-in-time retrospective focused on improving agent autonomy. Use when someone says "retro", "retrospective", "what did we learn", "what went wrong", "post-mortem", "incident review", or after a feature ships, an incident resolves, a pattern of agent failures emerges, or any event worth reflecting on. Analyzes what happened, identifies what blocked agent autonomy, and produces concrete improvements (learnings + tickets). Not calendar-driven — event-driven. The learning loop that makes the system self-improving.
npx claudepluginhub teambrilliant/marketplace --plugin tap-skillsHow this skill is triggered — by the user, by Claude, or both
Slash command
/tap-skills:retrospectiveThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Reflect on what happened. Identify what blocked agent autonomy. Produce concrete improvements so the system gets better.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Share bugs, ideas, or general feedback.
Reflect on what happened. Identify what blocked agent autonomy. Produce concrete improvements so the system gets better.
Not tied to sprints or calendars. Run a retro when there's something worth learning from: a feature shipped, an incident resolved, a pattern of failures, a project wrapped.
What happened that an agent couldn't handle autonomously, and what's the cheapest fix so it can next time?
What event prompted this retro? Determines scope and data to analyze.
Pull data relevant to the trigger:
git log --since="[relevant period]" → what was committed
gh pr list --state merged --search "[scope]" → PRs in scope
gh pr list --state closed --search "[scope]" → rejected PRs (signal!)
gh issue list --search "[scope]" → related issues
Also read if available:
.tap/system-health.md → system metrics around the event.tap/tap-audit.md → current readiness assessment.tap/learnings.md → prior learnings (avoid repeating known issues)Rejected PRs are gold. Every rejection is a gap in agent capability or context.
For each problem found, classify the root cause:
Context gap — agent lacked information to make the right choice
Harness gap — agent lacked tools or access to complete the task
Feedback gap — agent couldn't verify its own work
Design gap — the code itself was too complex to modify safely
Scope gap — agent took on work it shouldn't have
For each finding, write a concise learning:
[date] — [trigger]
- [what happened] → [root cause category] → [specific fix]
Good learnings are specific and actionable:
Bad learnings are vague:
Each learning that requires work becomes a ticket. Categorize by impact on agent autonomy:
Raises readiness score (MINIMAL→PARTIAL or PARTIAL→FULL):
Prevents repeat failures:
Reduces design complexity:
Human mode: Present findings. Propose tickets. Ask: "Which of these should we create issues for?"
Agent mode: Create issues via gh issue create with label retro-improvement. Human reviews and prioritizes.
Append (never overwrite) to .tap/learnings.md. Create the file and .tap/ directory if they don't exist.
Use the template in references/learnings-template.md.
Agents read .tap/learnings.md before starting work. Captured learnings prevent the same mistakes from repeating across sessions.