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:
- Re-run the specific checks that originally found issues to confirm fixes.
- Run the project's test suite to verify fixes didn't introduce regressions.
- Run build/compile to confirm no breakage.
- If new issues surfaced from fixes, add them to the fix queue.
- 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:
docs/dev-retrospective.md— Full development cycle analysisdocs/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
/iterateto start building a feature - Run
/researchto 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.mdin 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.mdso Phase 2 can read it. - Phase 2 MUST read the file written by Phase 1 from the
docs/directory. - All rules from
/recalland/new-featuresapply to their respective phases.