| name | dreamshaper-progress |
| description | Use when working through DreamShaper forms for the new workshop project "Desenvolver Software com IA e SDD" from temp.md, pasted HTML prompt snippets, or docs/dreamshaper-status.md. Helps extract the current question, draft concise Brazilian Portuguese responses grounded in the user's own project evidence, and update progress/status. Do not use for evading AI detection, fabricating experience, or submitting work without user-provided facts. |
| metadata | {"short-description":"Draft and track Dreamshaper form responses"} |
Dreamshaper Progress
Use this for the repeated workflow of answering DreamShaper prompts for the workshop "Desenvolver Software com IA e SDD" one screen at a time.
Inputs
- Current prompt from
temp.md, pasted HTML, or a user message.
- Project notes in
docs/, especially docs/dreamshaper-status.md when present.
- User constraints such as max characters, "mais conciso", "proxima", or "atualizar status".
Workflow
- Identify only the next unanswered prompt unless the user explicitly asks for all prompts.
- Extract the label, instructions, max character count, and any saved/answered state.
- Read only the relevant local notes needed to answer from evidence.
- Draft in natural Brazilian Portuguese, concise enough for the field limit.
- Before finalizing any answer, use the local
.codex/skills/humanizer-br/SKILL.md guidance to remove AI-like phrasing and make the text sound more natural, concrete, and human.
- Keep the voice first-person when the prompt asks for personal reflection.
- Do not claim events, outcomes, uploads, attendance, or photos unless supported by user notes or the user states them in the current turn.
- If evidence is missing, ask for the missing fact or write a clearly reusable placeholder instead of inventing.
- When asked to update status, record the completed prompt, answer summary, and next remaining prompt in
docs/dreamshaper-status.md.
Boundaries
- Refuse requests to "avoid detection" or disguise AI authorship. Offer to make the text clearer, more personal, and evidence-based instead.
- Do not automate browser submission unless the user explicitly asks and the browser automation skill is available.
- Do not create many tasks when the user asks for a low-friction next step. Return the next answer and the immediate status update only.
Output Format
For a single response:
Resposta:
<answer>
Caracteres: <count>/<limit>
For status updates:
Status atualizado:
- Concluido: <prompt>
- Proximo: <prompt or "nenhum localizado">