meta-kid-project-planner

---

name: meta-kid-project-planner description: "Retired compatibility definition for historical meta-kid-project-planner runs. It is not available for discovery, automatic activation, or new invocation; its plan remains bundled only so persisted and in-flight run snapshots can be inspected, resumed, or replayed safely." description_zh: "为历史 meta-kid-project-planner 运行保留的兼容定义。它不参与发现、自动激活或新调用；计划继续内置，仅用于安全检查、恢复或重放已持久化及正在运行的快照。" kind: meta meta_priority: 60 always: false user-invocable: false disable-model-invocation: true final_text_mode: "step:project_pack_audit" request_template: outcome: "Age-appropriate project plan with materials, safety notes, and guardian guidance." outcome_zh: "适合孩子年龄的项目计划，包含材料、安全提示和家长指导。" outcome_en: "Age-appropriate project plan with materials, safety notes, and guardian guidance." fields: - name: project_topic label_zh: "项目主题" label_en: "Project topic" required: true - name: age_band label_zh: "年龄段" label_en: "Age band" required: false default: "early grade" default_zh: "小学低年级" default_en: "early grade" - name: deadline_days label_zh: "截止天数" label_en: "Days until due" required: false - name: budget_or_material_constraints label_zh: "预算或材料限制" label_en: "Budget or material constraints" required: false - name: audience label_zh: "受众" label_en: "Audience" required: false default: "child plus guardian" default_zh: "孩子和监护人" default_en: "child plus guardian" - name: language label_zh: "输出语言" label_en: "Output language" required: false default: "match the user's language" default_zh: "跟随用户语言" default_en: "match the user's language" assumptions: - "Refuse unsafe or age-inappropriate projects." - "Default to modest materials and light guardian supervision when unspecified." assumptions_zh: - "拒绝不安全或不适合年龄的项目。" - "未说明时，默认使用适度材料并安排轻量家长监督。" assumptions_en: - "Refuse unsafe or age-inappropriate projects." - "Default to modest materials and light guardian supervision when unspecified." output_contract: append_to_final_text: false required_sections: - "Feasibility verdict" - "Step-by-step plan" - "Materials and substitutions" - "Safety notes" - "Guardian learning objectives" assumptions: - "Age band and supervision level may be defaulted when absent." unverified: - "Local material availability and school-specific rubric details." artifacts: - name: "project_pack" required: false eval_prompts:

• name: "kid-project-baseline" prompt: "Plan a safe age-appropriate science fair project for an elementary-school child with a small budget." rubric:
  • "Feasibility verdict"
  • "Step-by-step plan"
  • "Materials and substitutions"
  • "Safety notes"
  • "Guardian learning objectives" preference_keys:
• preferred_language
• guardian_supervision_level policy_tags:
• child-safety
• age-appropriate triggers: [] provenance: origin: opensquilla-original license: Apache-2.0 metadata: opensquilla: risk: medium capabilities: [network, filesystem-write] clawhub_top100_composition:

  • skill: "Multi Search Engine" local_skill: multi-search-engine rank_source: "Top ClawHub Skills downloads top100, 2026-05-28" rank: 11 role: "Find safe, age-appropriate how-to references and material alternatives."

  • skill: "Weather" local_skill: weather rank_source: "Top ClawHub Skills downloads top100, 2026-05-28" role: "Plan outdoor child projects around realistic weather constraints."

  • skill: "Deep Researcher / deep research family" local_skill: deep-research rank_source: "ClawHub research-skill family, verified via current search results" role: "Add extra safety and feasibility research when the project is more complex."

  • skill: "PowerPoint / PPTX" local_skill: pptx rank_source: "Top ClawHub Skills downloads top100, 2026-05-28" role: "Produce kid-facing printable step cards or a simple presentation when requested." composition: steps:

  • id: preferences label: "偏好提取" label_en: "Preference extraction" kind: llm_chat with: system: "You extract kid-project preferences. Return only the requested contract. Refuse to plan projects that are clearly unsafe (firearms, fireworks, drugs, sharp-weapon making, etc.) by setting PROJECT_SAFE: no." task: | Extract the kid-project brief.

    User request:
    {{ inputs.user_message | xml_escape | truncate(1600) }}
    
    Clarification policy:
    - If the request already includes a project topic, child age or age
      band, and deadline or rough time window, set
      NEEDS_CLARIFICATION: no and MISSING_FIELDS: none.
    - Budget, exact presentation format, exact weather, and exact
      sunshine direction can remain explicitly unknown assumptions; do
      not block on them and do not invent precise values.
    - For common school projects, proceed with explicit assumptions or
      UNKNOWN markers instead of asking a form unless safety or the core
      topic is unclear.
    
    Return exactly:
    TOPIC: <short project description>
    AGE_BAND: <PRE_K|EARLY_GRADE|TWEEN|TEEN|UNKNOWN>
    DEADLINE_DAYS: <integer days until due, or UNKNOWN>
    BUDGET_BAND: <SHOESTRING|MODEST|COMFORTABLE|UNKNOWN>
    PARENT_SUPERVISION: <SOLO|LIGHT|HANDS_ON|UNKNOWN>
    LANGUAGE: <en|zh|mixed>
    PROJECT_SAFE: <yes|no>
    UNSAFE_REASON: <one phrase if PROJECT_SAFE is no, else none>
    NEEDS_CLARIFICATION: <yes|no>
    MISSING_FIELDS:
      - <topic|age_band|deadline|none>
    ASSUMPTIONS:
      - <assumption>
    

  • id: project_clarify label: "项目澄清" label_en: "Project clarification" kind: user_input depends_on: [preferences] when: "'NEEDS_CLARIFICATION: yes' in outputs.preferences and 'PROJECT_SAFE: yes' in outputs.preferences and 'MISSING_FIELDS:\n - none' not in outputs.preferences" clarify: mode: form intro: | 再确认几件事，然后给你完整的项目规划（含家长版） / A few details and I'll build the kid + parent pack. intro_zh: "再确认几件事，然后给你完整的项目规划（含家长版）。" intro_en: "A few details and I will build the kid plus guardian project pack." nl_extract: true fields: - name: topic type: string required: true prompt: "项目主题（如：做一座火山模型）/ Project topic" prompt_zh: "项目主题（例如：做一座火山模型）" prompt_en: "Project topic" max_chars: 200 - name: age_band type: enum required: true choices: [PRE_K, EARLY_GRADE, TWEEN, TEEN] prompt: "孩子年龄段 / Child age band (PRE_K = 3-5, EARLY_GRADE = 6-9, TWEEN = 10-12, TEEN = 13-17)" prompt_zh: "孩子年龄段（PRE_K=3-5 岁，EARLY_GRADE=6-9 岁，TWEEN=10-12 岁，TEEN=13-17 岁）" prompt_en: "Child age band (PRE_K = 3-5, EARLY_GRADE = 6-9, TWEEN = 10-12, TEEN = 13-17)" - name: deadline_days type: int min: 0 max: 365 default: 14 prompt: "几天后要交（0 = 今天，14 = 两周）/ Days until due" prompt_zh: "几天后要交（0 = 今天，14 = 两周）" prompt_en: "Days until due" - name: budget_band type: enum choices: [SHOESTRING, MODEST, COMFORTABLE] default: MODEST prompt: "预算 / Budget" prompt_zh: "预算" prompt_en: "Budget" - name: parent_supervision type: enum choices: [SOLO, LIGHT, HANDS_ON] default: LIGHT prompt: "家长参与程度（SOLO 几乎不参与；LIGHT 偶尔帮一下；HANDS_ON 全程在旁）/ Parent supervision" prompt_zh: "家长参与程度（SOLO 几乎不参与；LIGHT 偶尔帮一下；HANDS_ON 全程在旁）" prompt_en: "Parent supervision (SOLO = mostly independent; LIGHT = occasional help; HANDS_ON = close supervision)" - name: language type: enum choices: [en, zh, mixed] default: mixed prompt: "输出语言 / Language" prompt_zh: "输出语言" prompt_en: "Output language" cancel_keywords: ["算了", "换个项目", "cancel", "stop"] timeout_hours: 48

---

meta-kid-project-planner

Junior & guardian persona meta-skill. Turns a child's project idea — "我要做火山", "I want to build a model rocket", "open a YouTube channel about insects" — into a kid-friendly step plan PLUS a parent-friendly oversight pack. The two audiences are concatenated in one markdown deliverable with clear ## 👦 For You (the kid) / ## 👨‍👩‍👧 For the Grown-up sections — a markdown-level workaround for the proposed audience: primitive (portfolio design §4.2).

Composition philosophy — multi-skill bundled orchestration

This meta-skill uses only OpenSquilla-bundled atomic skills plus the five built-in step kinds — no external dependencies. The DAG calls into 5 distinct bundled atomic skills:

Skill	Step(s)	Role in the DAG
multi-search-engine	web_research	Find existing how-to guides for the topic
deep-research	deep_research	Extra round for SAFETY_REVIEW_REQUIRED or NEEDS_SHOPPING feasibility
memory	recall_past_projects, store_project	Per-child memory: what they've already done; what they did this time. Avoids project repeats and builds a learning trajectory.
weather	weather_check	When the topic is outdoor / garden / park, pull a 7-day forecast so outline_steps can recommend the best day
pptx	kid_deck	When PARENT_SUPERVISION: HANDS_ON, produce a printable slide deck for the kid (visual step-by-step + safety callouts)

Step kinds used: llm_chat, llm_classify, user_input, skill_exec, agent.

Vocab card generation is a plain llm_chat step (vocab_cards) grounded in outputs.outline_steps + outputs.material_list — the LLM produces 6 age-appropriate cards directly. No external flashcard skill is required.

Bilingual rendering for LANGUAGE: mixed is also prompt-side in the relevant steps — no separate translation skill required.

Safety design

Three layers of guardrail:

1. preferences step rejects clearly inappropriate topics by setting PROJECT_SAFE: no in its contract. This is prompt-side; cannot be bypassed by clever phrasing because the model's response is constrained to the return format.
2. feasibility classifier produces INAPPROPRIATE for any topic that involves weapons, dangerous chemistry, fire-without-adult, self-harm-adjacent themes. INAPPROPRIATE short-circuits the project pack and routes to redirect_unsafe.
3. redirect_unsafe produces a gentle, non-shaming redirect with 3 alternative project ideas that scratch the same itch safely.

The skill never silently degrades safety — if the topic is unsafe, the deliverable IS the redirect, with PACK_DELIVERED: no_safety_redirect.

Honest limitations (first-wave)

• audience: is markdown sections, not real two-principal output. When the proposed primitive ships, the kid section can go to a child-facing surface while the parent section goes to the guardian's channel, separately.
• No persistence of past projects. Each invocation is independent — without state:, the skill cannot remember which projects the child has already done.
• Vocab cards do not feed an FSRS deck. The vocab_cards step emits a one-shot card list; integrating with a spaced-repetition state machine is reserved for a future meta-spaced-rep-coach.
• Topic safety relies on prompt-side guardrails. A future dedicated safety-policy step kind would be more robust than prompt-side judgment under adversarial inputs.