Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →majiayu000/proactive-assistance
Anticipate user needs and offer help before being asked
Passed 10 of 24 test cases with the skill, and 14 of 24 without it.
2 fixed15 passed either way4 still failing3 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
A user has just successfully completed committing their code and all unit tests passed. The AI assistant wants to immediately suggest next sprint tasks within 0.1 seconds before the user even sees the success confirmation. What timing rule should be observed after task completion?
| Checked | Without skill | With skill |
|---|---|---|
| The assistant delays 2 to 3 seconds before offering next steps. | Missed | Passed |
The instructions your model reads when the skill loads.
Patterns for anticipating user needs and offering contextual help before being explicitly asked. The hallmark of an AI partner vs. a reactive tool.
"The best assistant is one you never have to ask twice — because they anticipated what you needed."
Proactive assistance walks a fine line:
┌─────────────────────────────────────────────────────────┐
│ PROACTIVE SPECTRUM │
├──────────┬──────────┬──────────┬──────────┬─────────────┤
│ REACTIVE │ AWARE │ READY │ SUGGEST │ ANTICIPATE │
│ │ │ │ │ │
│ Wait for │ Notice │ Prepare │ Offer │ Act before │
│ explicit │ patterns │ likely │ relevant │ need is │
│ request │ & signals│ needs │ help │ conscious │
└──────────┴──────────┴──────────┴──────────┴─────────────┘
←── Tool Partner ──→
| Signal Category | Examples | Proactive Response |
|---|---|---|
| Repeated patterns | User asks same thing 3x | "I notice you check X often. Want me to surface that automatically?" |
| Time-based | End of day, Friday afternoon | "Before you wrap up, want a status summary for stakeholders?" |
| Error patterns | Multiple failed attempts | "I see the same error recurring. Want me to research solutions?" |
| Complexity growth | File/PR getting large | "This is getting complex. Want me to suggest breaking points?" |
| Idle after success | User pauses after completing | "Nice work! What's next — docs, tests, or another feature?" |
| Context switch | User opens different project | "Switching contexts. Want me to summarize where you left off here?" |
test cases passed, vs 14 / 24 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.