Goal Clarifier

Warm multi-turn goal clarification and action planning. Use when the user has a vague, oversized, or tangled goal and wants help thinking it through, narrowi...

Security Scan

The skill's requests and runtime instructions match its stated purpose (multi-turn goal clarification and planning); it is an instruction-only skill that uses only included reference files and system-provided context, and it does not ask for credentials or install anything.

Purpose & Capability

Name, description, and included reference documents all align with a conversational goal-clarification and planning skill. There are no unrelated required binaries, env vars, or config paths.

Instruction Scope

Runtime instructions stay within the stated purpose: they direct multi-turn clarification, limited questioning, periodic reflection, plan generation, and weekly-plan JSON output. The skill instructs the agent to use provided `[GOAL_CONTEXT]` data (when present) as background; this is coherent for progress-aware planning. All referenced files are bundled with the skill.

Install Mechanism

No install spec or external downloads; this is an instruction-only skill so nothing is written to disk or fetched at install time.

Credentials

The skill requires no environment variables or credentials. It uses system-provided context blocks (e.g., `[GOAL_CONTEXT]`, `[WEEKLY_CYCLE_REVIEW]`) which is appropriate for a progress-aware planner.

Persistence & Privilege

Flags are default (not always:true). The skill does not request permanent presence or attempt to modify other skills or system settings. Autonomous invocation is enabled by default (normal) but not excessive here.

Assessment

This is an instruction-only 'goal clarifier' skill that uses the bundled reference files and (when provided by the platform) system context blocks like `[GOAL_CONTEXT]` to reference task progress. It does not request credentials, install software, or call external endpoints in its instructions, so the attack surface is small. Before enabling: 1) confirm what your OpenClaw environment populates into `[GOAL_CONTEXT]` (it may include task names, notes, or other personal data) so you’re comfortable with the agent using that data to generate plans; 2) be aware the skill will emit structured JSON weeklyPlan objects (intended for automatic rendering/integration) — if downstream systems store or forward that JSON, treat its contents accordingly; 3) since the skill can be invoked autonomously (platform default), consider whether you want it auto-run in contexts where sensitive data might be present. Overall the package appears internally consistent and proportional to its purpose.

Turn fuzzy goals into realistic next steps.

References

• Chinese request: read ./references/guide-zh.md
• English request: read ./references/guide-en.md
• Full flow: ./references/workflow-zh.md or ./references/workflow-en.md
• Tone and sample outputs: ./references/examples-zh.md or ./references/examples-en.md
• Testing and iteration: ./references/eval-checklist-zh.md or ./references/eval-checklist-en.md

Rules

• Clarify before planning when the goal is vague, overloaded, conflicted, or unrealistic.
• Ask only 1-3 high-value questions per turn.
• Reflect back your understanding every 2-4 turns so the user feels heard and can correct course.
• Fit the plan to the user's real time, energy, budget, resources, dependencies, and execution style.
• Prefer a lighter plan the user can actually start over a complete but heavy plan.
• Stop clarifying once the key constraints and goal are clear enough; then switch into action planning.
• Keep the tone warm, structured, and natural. Do not sound like a form, interrogation, or therapy session.
• Respond in the user's language.
• Use the final output structure defined in the matching guide file.
• End with one grounded follow-up question that helps the user continue moving.
• When [GOAL_CONTEXT] data is provided in the message, use it to understand the current state of phases, tasks, and weekly plans. Reference specific completed/pending tasks by name when discussing progress or next steps. Never mention [GOAL_CONTEXT] tags to the user — treat this as background knowledge.
• After the initial roadmap is confirmed, transition naturally into weekly schedule planning. Ask about the user's daily available time, preferred time slots, and any recurring commitments before generating a detailed weekly plan.
• When a weekly plan cycle is ending or has ended, proactively suggest reviewing execution and planning the next week. Reference specific tasks that were completed or missed from the [GOAL_CONTEXT] data.