Agent skill · data analytics · aiskillstore
cross-task-learning
Pattern for aggregating insights across multiple tasks to enable data-driven evolution.
Why this skill is useful
Provides a structured approach to aggregate insights across tasks, enabling data-driven evolution that the AI wouldn't generate on its own.
What it needs
About 3k tokens when loaded. Last updated 2026-08-07. 405 stars on the source repository.
What this skill does
Cross-Task Learning Skill Pattern for maintaining aggregated insights across all completed tasks. When to Load This Skill Reflector: After writing individual reflection Evolver: Before analyzing reflections (to get aggregated view) Core Concept Individual reflections capture task-specific learnings. Cross-task learning aggregates these to identify: Patterns that keep appearing → Skill candidates Strategies that consistently work → Best practices Strategies that keep failing → Anti-patterns Bottlenecks that recur → System weaknesses Proposals that keep emerging → Priority improvements Aggregate File Location: memory/reflections/aggregate.json Example structure (compact JSON): Update Protocol (for Reflector) After writing individual reflection, update aggregate: Similarity Matching When checking if patterns are "similar": Thresholds for Action Metric Threshold Action -------- ----------- -------- Strategy effectiveness < 0.3 After 5 uses Flag as anti-pattern Strategy effectiveness > 0.8 After 5 uses Flag as best practice Failure pattern occurrences >= 3 Flag for urgent fix Skill candidate frequency >= 5 Propose as new skill Recurring discovery count >= 3 Add to knowledge base Recurring proposal count >= 3 Prioritize for evolution Query Patterns (for Evolver) Get top issues to fix: Get best practices to document: Get skill candidates ready for implementation: Get knowledge gaps: Integration with Evolver The evolver should: 1. Read aggregate.json FIRST (not individual reflections) 2. Use aggregated data for proposal prioritization: High-occurrence failure patterns → High priority High-frequency skill candidates → Medium priority Recurring proposals → Already validated ideas 3. Reference individual reflections only for details 4. Update recurringproposals[].status after evolution Principles 1. Aggregate, don't duplicate - Summary stats, not copies 2. Track trends - First seen, last seen, frequency 3. Enable queries - Structure for easy filtering 4. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aiskillstore/cross-task-learning