Scores task uncertainty to calibrate memory depth and iteration count. Use w hen stakes are high, facts are sparse, or user says quantify uncertainty, fa st slow think, uncertainty score. Triggers: quantify uncertainty, fast slow think,uncertainty score, how sure.
SkillScore82Maintained
Automatic
◆AI Engineer pack
Last verified 8/15/2026
What it changes
Passed 18 of 22 test cases with the skill, and 11 of 22 without it.
Without skill50%
With skill82%
Checks it passed that the model alone missed
Recommends the self-refine-loop mechanism
Invokes self-refine-loop to refine context
Assigns an uncertainty score greater than 7
and 11 more in the full report
9 fixed9 passed either way2 still failing2 broke
Pass-rate lift
+32 pp
Tokens
-22%
Agent turns
0%
Response time
-42%
Model gemini-3.6-flashMeasured 8/15/2026v218 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 9
Prompt
We are running standard npm test on a well-known repository with clear instructions, but we aren't sure if a non-critical test helper mock is outdated. Evaluate how sure you are.
What was checked, without and with the skill
Checked
Without skill
With skill
Assigns a procedural uncertainty score of 3 or lower
Missed
Passed
Assigns an epistemic uncertainty score of 4 or lower
Missed
Passed
case-19 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md36 linessingle file
Agentic Uncertainty Quantifier
When to Use
User says quantify uncertainty or task matches this capability
User says fast slow think or task matches this capability
User says uncertainty score or task matches this capability
User says how sure or task matches this capability
Workflow
Score epistemic uncertainty 0-10 (how much is unknown).
Score procedural uncertainty 0-10 (how clear are the steps).
High epistemic (>6): retrieve more context, run self-refine-loop.
Low procedural (<4): ask clarifying questions before acting.
Report scores and recommended depth to user.
Integrations
self-refine-loop
semantic-memory-manager
deep-search-enabler
Error Handling
Failure
Response
False confidence
Bias toward caution on destructive tasks.
Gotchas
Uncertainty > 7 on financial/deploy actions triggers hitl-approver.
Example
Input: User request matching triggers above.
Output: Structured result per workflow with integrations invoked as needed.