End-to-End Feature Repair
Trace a broken feature across files and dependencies, repair the root cause, and verify the full user flow.
Before & after
Same prompt, same agent — the only difference is the skill.
When you have a feature that's broken, it's important to understand what the problem is first. You might start by looking at error logs or trying to reproduce the issue. The problem could be in different parts of the system - maybe the data retrieval is slow, or the export formatting is inefficient. You could add some logging or monitoring to understand where the bottleneck is. Testing edge cases like large datasets helps identify scaling issues. Once you know the root cause, you can fix it systematically rather than trying random solutions.
FOCUSED FIX: EXPORT FEATURE RECOVERY (100K+ ROWS)
ROOT CAUSE ANALYSIS
Symptom: Timeouts on exports >100K rows, lost tail data
Investigation:
• Query logs show full table scan (no index on created_at)
• Memory profile: Export format loop holds full dataset in memory (3GB for 200K rows)
• Partial data loss: HTTP timeout at 60 seconds mid-export
PRIMARY ISSUE: Synchronous streaming without chunking
FIXED ISSUES
1. INDEXING: Add index on (user_id, created_at) for query speedup
→ Query time: 12s → 800ms for 1M-row scan
2. STREAMING: Replace in-memory buffer with chunked streaming
Old: Load 200K rows in memory → format → stream
New: Stream 10K row chunks → format → flush → repeat
→ Memory: 3GB → 150MB (99.7percent reduction)
3. TIMEOUT SAFETY: Increase HTTP timeout from 60s → 300s + heartbeat keepalive
→ Prevents connection drop mid-export
4. PARTIAL DATA: Add row counter + checksums to final export
→ Validates completeness before sending
TEST COVERAGE ADDED
✓ Export 10K rows (baseline)
✓ Export 100K rows (threshold)
✓ Export 500K rows (stress test)
✓ Simulate network timeout at 50percent export → should retry and resume
✓ Verify exported CSV matches source row count exactly
DEPLOYMENT PLAN
1. Add index (no downtime)
2. Deploy streaming code (feature flag: export_v2_streaming)
3. Canary: 10percent users on new export path
4. Monitor error rates, latency, completion rate
5. Ramp to 100percent over 48 hours
6. Cleanup: Remove old export code
VALIDATION METRICS
Before: 8percent timeout rate on 100K+ exports
After: <0.1percent timeout rate, 100percent data completeness
About this skill
“You fixed the error, but the feature still doesn't work.” That usually means the repair stopped at the visible symptom or missed a contract between components. This skill traces a named feature from its entry point to its observable result, fixes the failing boundary, and checks other consumers of the same logic.
What it does
- Maps the relevant execution path across handlers, shared helpers, storage, queues, workers, and external services.
- Separates observed failures from hypotheses and chooses checks that distinguish competing causes.
- Finds the earliest broken contract instead of hiding downstream errors with defaults or retries.
- Repairs the shared cause and inspects sibling callers for the same failure mechanism.
- Verifies the original reproduction, a related use case, and a relevant failure or retry path.
- Reports what was changed, what actually passed, and which parts remain unverified.
What you get
An implemented repair when code access and authorization are available, with a concise explanation of the causal chain and verification results. With pasted material alone, you get a provisional diagnosis, a concrete patch where supported, and the smallest checks needed to confirm it.
Best for
Developers repairing a feature that crosses multiple files or services, especially when earlier patches fixed one symptom and left the overall workflow broken.
SKILL.md preview
---
name: end-to-end-feature-repair
description: >-
Repair a broken feature or module by tracing its execution path, locating the earliest violated contract, fixing the shared cause, and verifying the original flow plus affected callers. Use when the user says "make notifications work end-to-end", "fix the checkout feature", "this module is broken", "the last patch didn't fix it", "my uploads keep getting stuck", "it works in the API but not the app", or "focus on getting account recovery working".
---
# End-to-End Feature Repair
… (sign up to view the full skill)More development skills
View all Development skills →Performance Bottleneck Profiling: Node, Python, Go
Classifies a slowdown, gives exact profiler commands, reads the output, and proves the fix with same-conditions p95/p99 numbers.
Feature Scaffolding with Working Tests
Turn a short feature spec into code, runnable tests, and integration wiring that match your existing project.
Feature Flag Rollouts and Kill Switches
Design staged flag rollouts, verify rollback paths, and audit flag debt with explicit defaults, gates, and retirement plans.
Microcontroller Firmware and RTOS Engineering
Build and debug MCU firmware with bounded ISR/RTOS timing, safe DMA ownership, and measured power budgets.