Bundled Claude Code skills

Ideafy ships three skills that teach the agent how to use Ideafy's MCP tools. Skills are markdown files loaded by Claude Code (or the equivalent path for Gemini / Codex) — they're not code, just instructions the agent reads before replying.

The three skills

`ideafy.md` — the main board interaction skill

This is the most-used skill. It instructs the agent to:

Detect the project on every invocation using mcp__ideafy__get_project_by_folder — stops if the agent isn't in an Ideafy-tracked folder
Parse the user's request for keywords: list, create, update, move, plan, test, opinion, pool, etc.
Execute the matching tool with parameters inferred from natural language
Report back with display IDs (e.g., KAN-42), titles, and any status transition that occurred

The skill enforces the display ID format ({prefix}-{taskNumber}) everywhere so you and the agent are always speaking the same card language.

`human-test.md` — card creation from a conversation

This skill is for the common flow where you've just had a conversation with the agent about a feature and you want to turn that conversation into a card, ready to test.

The skill extracts:

Title
Description (what the feature is)
Solution (how it was implemented)
Test scenarios, as checkbox lists grouped under ### Happy Path, ### Edge Cases, and similar H3 headings

It creates the card in progress status, then immediately calls save_tests to move it to Human Test. You end up with a populated card and a ticked-off solution tab, and the board reflects it in one round trip.

`product-narrative.md` — the narrative interview

The narrative interview is a seven-question flow that generates docs/product-narrative.md (or the path you configured). Questions cover story, problem, target users, core features, non-goals, tech stack, and success metrics. Before asking, the agent scans package.json, the README, and the top-level folder structure to ground its questions in your actual project.

The output is a markdown file with these sections, which the agent then loads as context for every evaluate run. See Your first project for the user-facing walkthrough.

How skills are installed

Click Install Ideafy Skills in the project settings modal. Under the hood this hits POST /api/projects/[id]/skills, which copies the three skill files into the project's .claude/skills/ directory.

You can verify via GET /api/projects/[id]/skills, which returns { installed: boolean }. You can remove them via DELETE.

The paths can be moved — Settings → General → Skills Path controls where Ideafy looks. The default is ~/.claude/skills (global) for Claude Code, or per-platform for Gemini/Codex. Project-local .claude/skills/ takes precedence when both exist.

Using skills from the card chat

In the card modal chat, type @skill and an autocomplete appears with all skills the agent can see — including the three bundled ones plus any you've added yourself. Pick one and the agent will load and apply it for the current turn.

See Card chat & mentions for the mention mechanism.

Writing your own skill

A skill is just a markdown file with YAML frontmatter:

---
name: my-skill
version: 1.0.0
description: What this skill does
---

# Instructions the agent will follow

...

Drop it into ~/.claude/skills/ (or the path you configured) and it's available via @skill. No restart required for new files in most agent CLIs — check your platform's behaviour if it doesn't show up.

Skills are an open-ended extension point. The three bundled ones are a starting point for common Ideafy operations, not an exhaustive list of what you can do.

Prev: Ideafy's MCP server Next: Bundled Claude Code hooks Up: User guide index