Facilitate workshop sessions in a one-step, multi-turn flow.
SkillScore82Maintained
Automatic
◆Project Management pack
Last verified 8/23/2026
What it changes
Passed 18 of 22 test cases with the skill, and 7 of 22 without it.
Without skill32%
With skill82%
Checks it passed that the model alone missed
Asks exactly one question.
Restates the pending turn question.
Marks one option with '(Recommended)'.
and 20 more in the full report
12 fixed6 passed either way3 still failing1 broke
Pass-rate lift
+50 pp
Tokens
+24%
Agent turns
0%
Response time
-26%
Model gemini-3.6-flashMeasured 8/23/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 12
Prompt
Ask me about our team's current deployment frequency in our DevOps workshop. I want quick choices to select from.
What was checked, without and with the skill
Checked
Without skill
With skill
Provides numbered answer options for selection.
Missed
Passed
Includes an 'Other (specify)' response option.
Missed
Passed
case-08 · gemini-3.6-flash · 8/23/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md70 linessingle file
Purpose
Provide the canonical facilitation pattern for interactive skills: one step at a time, with clear progress, adaptive recommendations at decision points, and predictable interruption handling.
Key Concepts
One-step-at-a-time: Ask a single targeted question per turn.
Session heads-up + entry mode: Start by setting expectations and offering Guided, Context dump, or Best guess mode.
Progress visibility: Show user-facing progress labels like Context Qx/8 and Scoring Qx/5.
Decision-point recommendations: Use enumerated options only when a choice is needed, not after every answer.
Quick-select response options: For regular context/scoring questions, provide concise numbered answer options plus Other (specify) when useful.
Flexible selection parsing: Accept #1, 1, 1 and 3, 1,3, or custom text, then synthesize multi-select choices.
Context-aware progression: Build on previous answers and avoid re-asking resolved questions.
Interruption-safe flow: Answer meta questions directly (for example, "how many left?"), restate status, then resume.
Fast path: If the user requests a single-shot output, skip multi-turn facilitation and deliver a condensed result.
Application
Start with a brief heads-up on estimated time and number of questions.
Ask the user to choose an entry mode:
1 Guided mode (one question at a time)
2 Context dump (paste known context; skip redundancies)
3 Best guess mode (infer missing details and label assumptions)
Run one question per turn and wait for an answer before continuing.
Keep questions plain-language; include a short example response format when helpful.
Show progress each turn:
Context Qx/8 during context collection
Scoring Qx/5 during assessment/scoring
Ask follow-up clarifications only when they materially improve recommendation quality.
For regular context/scoring questions, offer quick-select numbered response options when practical:
Keep options concise and mutually exclusive when possible.
Include Other (specify) if likely answers are open-ended.
Accept multi-select responses like 1,3 or 1 and 3.
Provide numbered recommendations only at decision points:
after context synthesis,
after maturity/profile synthesis,
during priority/action-plan selection.
Accept numeric or custom choices, synthesize multi-select choices, and continue.
If interrupted by a meta question, answer directly, then restate progress and pending question.
If the user says stop/pause, halt immediately and wait for explicit resume.
End with a clear summary, decisions made, and (if best guess mode was used) an Assumptions to Validate list.