Clarity First

Intent detection protocol for Claude — identifies the real goal behind requests, surfaces hidden assumptions, and knows when to ask vs. when to proceed. Elim...

Security Scan

The skill's instructions and metadata align with its stated purpose (intent detection/clarification); it requires no credentials, no installs, and contains no surprising behavior.

Purpose & Capability

Name and description (intent detection / clarification) match the SKILL.md: the protocol describes translating intent, listing assumptions, scoring ambiguity, and guarding scope. No unrelated permissions, binaries, or env vars are requested.

Instruction Scope

Runtime instructions are limited to restating requests, enumerating assumptions, asking up to three focused questions, and declaring scope. They do not reference filesystem paths, environment variables, external endpoints, or any data-exfiltration steps.

Install Mechanism

This is an instruction-only skill with no install spec and no code files — nothing is written to disk or downloaded during install.

Credentials

The skill declares no environment variables, credentials, or config paths. There are no unexpected secrets requested that would be disproportionate to intent clarification.

Persistence & Privilege

Flags are default (always: false, user-invocable: true, agent invocation allowed). There is no request for permanent presence or modification of other skills/config; privilege level is appropriate.

Assessment

This skill is coherent and low-risk: it only instructs the agent to ask clarifying questions and state assumptions. Before installing, decide whether you want the agent to routinely pause and ask questions (it can change workflow cadence). Monitor the first few runs to ensure clarifying questions do not inadvertently prompt the user to reveal sensitive secrets, and be cautious if you later combine this with other skills that perform networked actions (e.g., task executors) — the clarification protocol itself does not perform any external operations.

Don't execute the wrong thing perfectly. Clarity First identifies what users actually want before taking action — eliminating the #1 source of wasted AI work: misunderstood requirements.

The Core Problem

Users say what they ask. They mean something slightly different. The gap causes rework.

• "Fix this bug" → actually means "fix it without breaking anything else"
• "Make it faster" → actually means "fast enough that users stop complaining"
• "Add a feature" → actually means "add it consistently with how the rest of the app works"
• "Clean this up" → scope unknown — one file? the whole codebase?

When to Activate

Use Clarity First before:

• Starting any new feature or significant change
• Interpreting an ambiguous or multi-solution request
• Taking an action that is hard to reverse
• Receiving a request that could be fulfilled in meaningfully different ways

Skip it for: simple factual questions, single-step operations, requests already handled the same way earlier in the session.

The Protocol

Step 1: Intent Translation

Before doing anything, translate the literal request into the real goal:

Said:            "..."
Means:           "..."
Success looks like: "..."

If the translation differs from the literal request, flag it. Ask if the translation is correct before proceeding.

Step 2: Assumption Inventory

List every assumption required to fulfill the request. Be specific:

• Technical — language, framework, runtime, version, environment
• Scope — what is in and out of bounds
• Quality — how good is "good enough"; performance/test/style bar
• Constraints — backwards compatibility, deadlines, existing patterns to follow

Step 3: Ambiguity Score

Count the number of critical unknowns — things where the wrong assumption causes rework:

Unknowns	Action
0–1	Proceed. State your assumptions inline.
2–3	Ask the single most important question. State the rest as assumptions.
4+	Ask up to 3 focused questions before starting.

Never ask more than 3 questions at once. Prioritize ruthlessly.

Step 4: Scope Guard

Before executing, state the scope boundary explicitly:

In scope:  ...
Out of scope: ...

If the user expands scope mid-task, pause and re-run the protocol for the new scope.

Output Format

For non-trivial requests, open with a brief Clarity Check:

[Clarity Check]
You want: ...
I'm assuming: ...
Confidence: high / medium / low
→ Proceeding / → One question first: ...

Keep it short. The Clarity Check should take 3 lines, not 3 paragraphs.

Pairs Well With

• thinkdeep — analyze the solution after the problem is well-defined
• task-pilot — create an execution plan once requirements are clear

Install the full ThinkStack for best results:

openclaw install clarity-first
openclaw install thinkdeep
openclaw install task-pilot