Ideas Notes

Ideas Notes Skill

Use this skill whenever the user asks to:

• capture a new idea;
• revisit prior ideas;
• evolve a proposal into actionable steps;
• check idea status/history before implementation.

This skill must also support:

• listing all registered ideas;
• fetching a specific idea by ID or by topic keywords;
• producing concise and executive summaries of one or many ideas.

Collaboration mode (mandatory)

• The agent must actively help the user evolve ideas while discussing them.
• Do not act as a passive note taker only.
• During idea conversations, the agent should:
  • propose stronger alternatives;
  • suggest simpler implementations when possible;
  • suggest phased rollout options (MVP -> incremental);
  • compare trade-offs (effort, risk, usability, maintainability);
  • convert abstract ideas into actionable next steps.
• If user idea has gaps, propose concrete improvements and ask focused questions.
• If a better solution is identified, explicitly present it with rationale.

Trigger phrases (mandatory dispatch)

Whenever user writes idea-intent phrases (or close synonyms), dispatch this skill first.

Examples of trigger phrases:

• "tive uma ideia"
• "tenho uma ideia"
• "pensei numa ideia"
• "me veio uma ideia"
• "vamos anotar uma ideia"
• "quero registrar uma ideia"

After a trigger, the agent must ask this disambiguation question before proceeding:

• "Isso é continuidade de alguma ideia existente ou é uma ideia nova?"

Then map user answer to intent:

• existing idea -> get or update
• new idea -> create

Source of truth

• Idea index: ~/.copilot/skills/ideas-notes/notes/index.md
• Idea files: ~/.copilot/skills/ideas-notes/notes/*.idea.md
• Idea template: ~/.copilot/skills/ideas-notes/notes/template.idea.md

Required workflow

1. Read ~/.copilot/skills/ideas-notes/notes/index.md first.
2. Resolve user intent:
   • list: user wants all ideas;
   • get: user wants one specific idea (by ID/topic);
   • summary: user wants a compact synthesis;
   • create: user wants a new idea;
   • update: user wants to evolve an existing idea.
   • if trigger phrase was used and user did not clarify, ask the mandatory disambiguation question first.
3. If intent is list:
   • return the idea table with ID, topic, status, priority, last update;
   • if requested, filter by status/priority/topic.
4. If intent is get:
   • search by explicit ID first (IDEA-XXX), then by topic keywords;
   • open the matching .idea.md file and return current status, latest revision, next step.
5. If intent is summary:
   • summarize by domain and status;
   • highlight risks, blockers, and suggested next action.
6. If intent is create:
   • create a new ID (IDEA-XXX);
   • create a new file from template.idea.md named like:
     • idea-001-short-topic.idea.md;
   • add row in index.md linking the new file.
7. If intent is update:
   • update status/date in index.md;
   • add a NEW revision entry at the TOP of the idea file revision log.

Critical analysis policy (mandatory)

• Do not accept proposals blindly.
• Challenge assumptions with clear technical reasoning when needed.
• Always call out potential:
  • scope leaks;
  • side effects;
  • security risks;
  • usability risks;
  • maintainability concerns.
• Ask clarifying questions when requirements are ambiguous or contradictory.
• When rejecting or warning, provide safer alternatives and trade-offs.

Internal tools policy

• Skill may create helper scripts for agent-only usage under:
  • ~/.copilot/skills/ideas-notes/tools/
• Tools catalog file (mandatory):
  • ~/.copilot/skills/ideas-notes/tools/index.md
• Allowed extensions:
  • .tool.js, .tool.py, .tool.sh
• Example:
  • ~/.copilot/skills/ideas-notes/tools/read-vue-file.tool.py
• Purpose of tools:
  • inspect project folders/files quickly;
  • extract/minify code for faster agent consumption;
  • build derived artifacts (maps/graphs/indexes) for idea analysis.
• Rules for tools:
  • tools are for agent workflow, not user-facing features;
  • always consult tools/index.md before reading tool source files;
  • prefer creating tools only when reuse value is clear;
  • keep tools deterministic and scoped;
  • document tool purpose in file header comment;
  • never print secrets or credentials.

Rules

• Keep content in English unless the user explicitly asks another language.
• Never delete historical ideas; mark as blocked, needs-improvement, or completed when needed.
• Keep IDs stable once created.
• Keep statuses limited to:
  • draft, planned, in-progress, validated, needs-improvement, blocked, completed.
• Never rewrite old revision entries in .idea.md files.
• Revision history must be sequential and newest-first.
• Expected order over time:
  • {datetime} - conclusion
  • {datetime} - rev-02
  • {datetime} - rev-01
  • {datetime} - new-idea

Output style

• Concise summary + clear next actions.
• Always reference idea IDs.
• For critical feedback, include: risk, impact, and mitigation.