Skill

Park Idea Skill

Quick idea capture while working - park thoughts without breaking flow. Use when user says: "park this idea", "park this thought", "save this for later", "side thought", "note for later", "idea for later", "don't forget", "remind me later", "come back to this", "tangent:", "btw:", "also:"

From task-decomposer
Install
1
Run in your terminal
$
npx claudepluginhub jskswamy/claude-plugins --plugin task-decomposer
Tool Access

This skill uses the workspace's default tool permissions.

Skill Content

Park Idea Skill

You help users quickly capture side thoughts and ideas without breaking their flow. The goal is minimal interruption - get the idea captured and return to work immediately.

Workflow

1. Detect Current Context (Automatic)

Gather context silently without asking the user:

# Find current in-progress task
bd list --status=in_progress --json 2>/dev/null | head -1

Also note from the conversation:

  • What file/code we were discussing
  • What triggered this idea (the recent topic)

2. Capture the Idea

Ask only one question:

What's the idea?

If the user already provided the idea in their message (e.g., "park this - we should add caching"), extract it directly without asking.

3. Create Parked Issue

bd create "PARKED: {brief title from idea}" \
  -t task \
  -p 4 \
  --status deferred \
  -l parked-idea \
  --description "## Idea
{user's idea text}

## Context
Parked while working on: {current-task-id or 'no active task'}
Triggered by: {what we were discussing}
File context: {current file if any, or 'N/A'}"

If there's a current in-progress task, also add a dependency link:

bd dep add {new-idea-id} {current-task-id}

This creates a "discovered-from" relationship.

4. Confirm Briefly

Keep confirmation minimal:

Parked as {id}! Continuing with {current task or 'your work'}...

Then immediately return focus to what the user was doing before.


Key Principles

  1. Maximum 1 question - If the idea is clear, ask nothing
  2. Auto-capture context - Don't ask "what task are you working on?"
  3. Use deferred status - Parked ideas don't clutter active work
  4. Label consistently - parked-idea label enables easy filtering
  5. Link to source - Dependency shows where idea came from
  6. Return to work fast - The whole interaction should take seconds
  7. Beads is the memory - All parked ideas live in beads, not in markdown files or framework-specific storage. This ensures they survive across sessions, compaction, and context limits. No matter which decomposition framework the project uses, parked ideas are always beads issues.

Examples

User provides idea inline

User: "btw: we should probably add rate limiting to this API"

Response:

bd create "PARKED: Add rate limiting to API" -t task -p 4 --status deferred -l parked-idea --description "..."

Parked as claude-plugins-abc! Continuing...

User wants to park something

User: "park this thought"

Response:

What's the idea?

User: "The config validation could use JSON schema instead of manual checks"

Response:

bd create "PARKED: Use JSON schema for config validation" ...

Parked as claude-plugins-xyz! Back to the auth refactor...


Finding Parked Ideas Later

Users can review parked ideas with:

  • /review-parked - dedicated review skill
  • bd list --status=deferred -l parked-idea - direct command
  • Automatic prompt after task commits (via hook)
Stats
Parent Repo Stars1
Parent Repo Forks0
Last CommitMar 18, 2026