Identify repeated workflows, recurring user asks, common failures, and reusable procedures that should become new skills or improvements to existing skills....
name: skill-learner
description: Identify repeated workflows, recurring user asks, common failures, and reusable procedures that should become new skills or improvements to existing skills. Use when patterns repeat, when the user corrects the same kind of mistake, or when a task keeps requiring the same instructions, docs, or tool flow.
Skill Learner
Overview
Use this skill to turn repeated work into reusable skill knowledge. Focus on extracting patterns, not doing the underlying task again.
Learning workflow
1. Identify the pattern
Look for repeated signals such as:
the same type of task happening multiple times
the same correction from the user more than once
repeated tool sequences
repeated failures or omissions
repeated project workflows that should be standardized
2. Decide the right outcome
Choose one:
create a new skill
improve an existing skill
add a reference file or script to an existing skill
record a smaller local operating rule instead of creating a full skill
3. Extract the reusable pieces
Capture only what is worth reusing, such as:
triggers for when the skill should be used
the core workflow
decision rules
scripts or references that would prevent repetition
common pitfalls and corrections
4. Recommend the smallest useful skill shape
Prefer the lightest structure that solves the repeated problem:
SKILL.md only
SKILL.md + references
SKILL.md + scripts
5. Report clearly
Use this structure:
Pattern observed
what keeps repeating
Why it should become skill knowledge
what waste/error it prevents
Recommended action
new skill / improve existing skill / local rule only
Minimal skill shape
what files/resources are actually needed
Rules
Do not create a skill unless the pattern is real and reusable.
Prefer improving an existing skill over creating duplicates.
Keep recommendations lean.
Avoid auxiliary documentation clutter.
If a local workspace rule is enough, say so instead of forcing a new skill.
π Constraints
Do not create a skill unless the pattern is real and reusable.
Prefer improving an existing skill over creating duplicates.
Keep recommendations lean.
Avoid auxiliary documentation clutter.
If a local workspace rule is enough, say so instead of forcing a new skill.
