askme

name

askme

description

Verbalized Sampling (VS) protocol for deep intent exploration before planning, now mode-aware. Default `exhaustive` mode runs the existing VS protocol verbatim (callers without a mode arg get unchanged behavior). Optional `collaborative` mode runs a two-way tip-sharing dialogue; optional `adversarial` mode walks the design tree one fork at a time with recommendations per question. Mode-selection is hybrid — auto-detect from invoking-context phrasing ("help me refine" → collaborative, "poke holes" → adversarial, otherwise exhaustive) with explicit override via `/askme adversarial|collaborative|exhaustive`. Use when starting ambiguous or complex tasks, when multiple interpretations exist, or when you need to explore diverse intent hypotheses and ask maximum clarifying questions before committing to an approach.

Ask Me Command

Before proceeding to ask planning questions, you must proactively and critically execute both Verbalized Sampling (VS) and exploration:

• For Verbalized Sampling, generate and sample at least N distinct, diverse candidates that represent different possible user intents or directions, ranked by likelihood, where N is dynamic by ambiguity/risk/scope (baseline N>=5; trivial N>=3; high ambiguity/risk N>=7; architectural N>=10; no hard cap). Run actor-critic on each VS sample: explicitly record one weakness, contradiction, and oversight before selecting a direction. VS prevents over-engineering by surfacing simpler alternatives; expand only while new samples materially change planning decisions, and prefer the smallest sufficient N.

Required VS Output Format:

1. [Most likely] hypothesis here
   - Weakness: [potential flaw]
   - Contradiction: [logical conflict if any]
   - Oversight: [what this misses]

2. [Alternative] hypothesis here
   ...

• For exploration, deliberately seek out unconventional, underexplored, and edge-case possibilities relating to the user's objective, drawing on both the provided context and plausible but non-obvious requirements. Include at least 3 edge cases (at least 5 if architectural), and stop expanding once additional cases no longer change decisions.

Only after completing both critical VS and exploration steps, proceed to use the question tool to ask the maximum possible number of precise, clarifying, and challenging planning questions that holistically address the problem space, taking into account uncertainty, gaps, and ambiguous requirements.

Modes [LOAD-BEARING]

askme is intentionally one skill with three modes (not split into per-mode skills); callers without a mode arg get exhaustive (the original VS protocol) unchanged. The body above describes exhaustive; the two new modes share the AskUserQuestion contract and antipattern guidance below but differ in posture.

Mode-selection (hybrid)

Auto-detect from invoking-context phrasing, with slash-arg override:

• User wording matches help me refine, walk through with me, let's brainstorm, share tips → collaborative.
• User wording matches poke holes, stress-test, grill, find weaknesses → adversarial.
• Anything else, including no qualifier → exhaustive (current VS behavior; backward-compatible default).
• Explicit override: /askme exhaustive, /askme collaborative, /askme adversarial. The override always wins over auto-detect. Callers that invoke without any mode arg get exhaustive — zero behavior change for existing invocations.

exhaustive mode

The VS-shaped protocol described above. Sample N intent hypotheses, rank, run actor-critic on each, then fire the maximum-cardinality clarifying question batch. This is the default when no other signal fires.

collaborative mode

Two-way tip-sharing dialogue. Surface one of your own observations / tips back to the user as a counter-tip per round; let depth emerge through exchange. No scoring, no ranked sample. Use when the user is exploring a problem space rather than approaching commitment, or when their wording explicitly invites collaboration. Stop when the user signals convergence.

adversarial mode

Walk the design tree one fork at a time. Per fork: state the question, recommend an answer with one-sentence rationale, wait for the user, do not proceed on assumed answers. Resolve dependencies parents-first; do not descend into children while a parent is unresolved. Stop conditions: every fork has a committed answer, a blocking unknown surfaces, or the design dissolves under questioning. Adjacent to grill-me (general design grilling) and grill-ai-mastery (AI-vocabulary anchored grilling) — pick those when the design under test is the anchor; pick this mode of askme when intent is the anchor and adversarial probing is what the user asked for.

Escalation triggers (collaborative → adversarial)

Promote from collaborative to adversarial mid-session when any of these fire:

• Ambiguity cardinality ≥ 2 valid architectural decisions surface from a single user message; force a fork.
• Unspecified file paths — the user references "the function" / "that file" without a concrete path; demand resolution before continuing.
• Missing success criteria — the user describes a goal without a verifiable signal of done; surface the gap.

Sources: signal patterns documented in Cursor Plan Mode best practices and the NeurIPS 2025 Multi-Agent Clarification (MAC) paper. Treat as guidance, not a hard rule — the user can always override mode via slash-arg.

Mode interactions with the VS preamble

The VS preamble (sample N, rank, actor-critic) is required only in exhaustive mode. collaborative does not run VS — it foregrounds dialogue. adversarial runs VS once at the start to map the design tree, then proceeds per-fork.

AskUserQuestion tool contract (Claude Code reference)

This protocol assumes a single "ask user" tool with the contract below. Other agent harnesses (Codex, Gemini CLI, Aider, OpenAI Assistants, …) should map their equivalent question/prompt tool to this surface — field names and numeric limits below are Claude Code's AskUserQuestion; the shape is what the protocol depends on, and the (Recommended) convention is what the per-axis pick semantics rest on.

Antipattern: override-checklist UI [LOAD-BEARING]

Bad shape — never generate this:

Which of these defaults should I override before I lock in the plan?
❯ 1. [ ] Diff-only mode
  2. [ ] Include root prompts
  3. [ ] system-prompt-baseline.md wins on conflict
  4. [ ] Bump plugin manifests

This is a single multiSelect: true question where unticked = "default stands". It collapses four independent axes into one checkbox list. Never generate this shape.

Correct shape — one single-select question per axis:

Q1 — Scope (single-select)
❯ Diff-only mode (Recommended) — propagate only recently-new baseline
  Full block alignment — full sweep across all blocks

Q2 — Roots (single-select)
❯ Skip root prompts (Recommended) — derivative artifacts
  Include root prompts — also touch ODD/{GENERIC,COMPACTED,MINIMAL}

Q3 — Conflict policy (single-select)
❯ Preserve target policy (Recommended) — non-conflicting only
  system-prompt-baseline.md wins — override divergent target policy

Q4 — Manifests (single-select)
❯ Skip bump (Recommended) — sibling-harness scope
  Bump minor — semver per system-prompt-baseline.md memory note

Positive routing rule: When the brief calls for the user to rarely have to type, route the intent into N per-axis single-select questions (≤4 per fire) — each axis's (Recommended) option carries the default. Ticking (Recommended) is accepting the default.

Never use multiSelect for axis-with-default override semantics. Reserve multiSelect strictly for additive picks (feature toggles, optional sub-tasks).

Per fire (one tool call):

• questions array — minItems: 1, maxItems: 4. All questions in the array render as one batched UI; one user round-trip per fire.

Per question:

• question — full sentence ending in ?
• header — short chip label, ≤ 12 characters
• multiSelect — boolean (default false). false = single-pick (mutually exclusive options); true = subset of additive items (feature toggles, optional sub-tasks)
• options — array, minItems: 2, maxItems: 4

Per option:

• label — 1-5 words; the chip text the user sees and ticks. Mark the recommended choice by appending (Recommended) to its label and placing it first in the array.
• description — explanation of the trade-off / consequence; the one-sentence rationale lives here.
• preview — optional rendered content (markdown, monospace box). Single-select only (tool constraint). Use for visual comparisons (layout mockups, code diffs, file trees); skip when the difference is purely conceptual.

Built-in escapes (do not duplicate):

• The free-text "Other" input is auto-provided on every question; never add an explicit "Other" option.
• Users may attach free-text notes via the annotations response field.

Plan-mode caveat:

• Use this tool only to clarify requirements or choose between approaches during planning. Do not ask "Is the plan ready?" / "Should I proceed?" — that's what ExitPlanMode is for.

askme-specific notes:

• The "maximum possible number" of questions above is bounded by the tool's per-fire cap (4); for larger question sets, fire multiple sequential batches, ordered by dependency.
• Render the VS block immediately before the first AskUserQuestion fire of a planning session; subsequent intra-session fires need not repeat the VS preamble unless the survivor set materially changed.

Mapping for other harnesses:

• If the harness exposes only single-question prompts, fire them sequentially in the dependency order — the shape (clarifying questions with one Recommended each) is what matters; batching is an optimization.
• Map (Recommended) to whatever default-marker convention the harness uses; the rationale belongs in the description body either way.
• Map multiSelect: true to whatever multi-pick mechanism the harness exposes; if none, decompose additive picks into N independent single-selects.