agentsclimarketplace

Retro

Skill tinh2/skills-hub-registry/combo/retro

Run a full retrospective and dev cycle analysis. Chains /recall → /new-features to reconstruct what went wrong, extract lessons learned, identify rework patterns, and synthesize feature ideas from the findings. Triggers: retrospective, retro, dev cycle analysis, what went wrong, lessons learned, postmortem, sprint retro.From its SKILL.md

Install
npx -y skills add tinh2/skills-hub-registry --skill retro

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 12 stars12 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.7 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

You are an autonomous retrospective-to-ideation agent. Do NOT ask the user questions. Run the full pipeline below without pausing between phases.

TARGET: $ARGUMENTS

If arguments are provided, use them to scope the retrospective (branch name, date range, feature). If no arguments are provided, analyze the full git history.

============================================================ PHASE 1: DEVELOPMENT RETROSPECTIVE (/recall)

PARALLEL EXECUTION: After /recall completes and produces findings, use the Agent tool to run follow-up analyses concurrently.

  • Agent A (Feature Discovery): "Based on these recall findings: [findings summary], run /new-features to discover features that address the identified bottlenecks and rework patterns."
  • Agent B (Metrics Check): "Based on these recall findings: [findings summary], compute development quality metrics and compare to baseline."
  • Wait for both agents to complete and merge into the retrospective output.

Follow the instructions defined in the /recall skill exactly.

Produce the full retrospective output: Timeline Reconstruction, Pattern Extraction, Dependency Mapping, and Insight Distillation.

Save the retrospective output to docs/dev-retrospective.md (create the docs/ directory if it doesn't exist).

Key outputs to preserve for Phase 2:

  • What caused rework (indicates areas needing better features/tooling)
  • Where bottlenecks were (indicates areas needing improvement)
  • Patterns of late discovery (indicates missing validation/features)
  • Recommended improvements for the next cycle

Do NOT stop here. Continue immediately to Phase 2.

============================================================ PHASE 2: FEATURE IDEATION (/new-features)

Now follow the instructions defined in the /new-features skill exactly.

Read ALL .md files in the docs/ directory — this includes the docs/dev-retrospective.md you just wrote plus any other markdown files that already existed in the folder.

Extract learnings from every file, synthesize feature ideas, and write the report to docs/NewFeatures-X.md (where X is a 3-word kebab-case theme).

IMPORTANT: Features should directly address the pain points and inefficiencies discovered in the retrospective:

  • Rework hotspots → features that prevent those issues
  • Bottlenecks → features that eliminate or reduce them
  • Late discoveries → features that catch issues earlier
  • Process gaps → features or tooling that fill them

============================================================ SELF-HEALING VALIDATION (max 3 iterations)

After completing all phases, validate the combined output:

  1. Re-run the specific checks that originally found issues to confirm fixes.
  2. Run the project's test suite to verify fixes didn't introduce regressions.
  3. Run build/compile to confirm no breakage.
  4. If new issues surfaced from fixes, add them to the fix queue.
  5. Repeat the fix-validate cycle up to 3 iterations total.

STOP when:

  • Zero Critical/High issues remain
  • Build and tests pass
  • No new issues introduced by fixes

IF STILL FAILING after 3 iterations:

  • Document remaining issues with full context
  • Classify as requiring manual intervention or architectural changes

============================================================ OUTPUT

When both phases are complete, print a summary:


Retro Complete

Files generated:

  1. docs/dev-retrospective.md — Full development cycle analysis
  2. docs/NewFeatures-[X].md — Feature ideas derived from retrospective learnings

Key stats:

  • Commits analyzed: [N]
  • Rework hotspots identified: [N]
  • Feature ideas generated: [N] (HIGH: [N], MEDIUM: [N], LOW: [N])

Top insight: [single most impactful finding from the retrospective]

Next steps:

  • Run /spec [feature name] to generate implementation stories
  • Run /iterate to start building a feature
  • Run /research to validate features against competitors

============================================================ SELF-EVOLUTION TELEMETRY

After producing output, record execution metadata for the /evolve pipeline.

Check if a project memory directory exists:

  • Look for the project path in ~/.claude/projects/
  • If found, append to skill-telemetry.md in that memory directory

Entry format:

### /retro — {{YYYY-MM-DD}}
- Outcome: {{SUCCESS | PARTIAL | FAILED}}
- Self-healed: {{yes — what was healed | no}}
- Iterations used: {{N}} / {{N max}}
- Bottleneck: {{phase that struggled or "none"}}
- Suggestion: {{one-line improvement idea for /evolve, or "none"}}

Only log if the memory directory exists. Skip silently if not found. Keep entries concise — /evolve will parse these for skill improvement signals.

STRICT RULES:

  • Do NOT skip Phase 1 and go straight to Phase 2.
  • Do NOT ask the user for input between phases.
  • Phase 1 MUST save its output to docs/dev-retrospective.md so Phase 2 can read it.
  • Phase 2 MUST read the file written by Phase 1 from the docs/ directory.
  • All rules from /recall and /new-features apply to their respective phases.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.