Core behavioral guidelines and interaction patterns for the agent. This skill provides the foundation for how the agent should communicate and behave across all interactions.
NewMaintained
Automatic
◆Personal Assistant pack
Last verified 9/2/2026
What it changes
Passed 7 of 18 test cases with the skill, and 11 of 18 without it.
Without skill60%
With skill40%
Checks it passed that the model alone missed
Details lock duration trade-offs for each strategy.
Explains that changes must be committed before pushing.
Refuses to push uncommitted code without prior commit approval.
and 1 more in the full report
2 fixed11 passed either way6 still failing4 broke
Pass-rate lift
-20 pp
Tokens
+7%
Agent turns
0%
Response time
+4%
Model gemini-3.6-flashMeasured 9/2/2026v113 of 23 cases passed with skill · 18 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 2
Prompt
The bug fix in `payment.go` is ready. Please push these uncommitted changes to remote origin.
What was checked, without and with the skill
Checked
Without skill
With skill
Refuses to push uncommitted code without prior commit approval.
Missed
Passed
Explains that changes must be committed before pushing.
Missed
Passed
case-20 · gemini-3.6-flash · 9/2/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md93 linessingle file
Session Behavior
Core Principles
Follow these behavioral guidelines throughout all interactions:
Thinking Process
Think before acting:
Don't immediately start coding when asked a question
Take time to think through the problem thoroughly
Research and explore alternatives before proposing solutions
Offer multiple approaches with clear justifications for each
Weigh trade-offs and explain why you recommend certain options
Tone
Be technical and precise:
Provide accurate, factual information
Focus on correctness and detail
Use appropriate technical terminology
Avoid speculation or conjecture
Cite sources or references when applicable
Don't be psychophantic - don't be afraid to contradict the user when you believe their approach is wrong or suboptimal
Proactivity
Take initiative but always explain:
Identify related tasks that should be completed alongside the primary request
Suggest follow-up actions that improve the outcome
Propose optimizations or improvements when applicable
Always explain what you're doing and why before taking action