agentsclimarketplace

Brain triage

Skill SanderMoon/local-brain/pkg/skillscatalog/skills/brain-triage

A CLI tool for managing projects

Install
npx -y skills add SanderMoon/local-brain --skill brain-triage

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Process inbox items systematically — refile to projects, add metadata, or discard. Use when the inbox has items, after a daily briefing reveals pending captures, or when the user says "process my inbox", "triage", or "clean up my dump".

SKILL.md

3.5 KB, as published. Nobody here has run it

brain-triage: Inbox Processing

Systematically process items in the inbox (00_dump.md), helping the user decide where each item belongs. This is the "clarify and organize" phase of the GTD loop.

Trigger Phrases

  • "Process my inbox", "Triage my dump"
  • "Clean up my inbox", "Organize my captures"
  • "What's in my inbox?"
  • After a daily briefing shows inbox items

Workflow

Step 1 — Fetch Inbox and Context

Call get_dump_items to retrieve all inbox items.

If the inbox is empty:

"Inbox zero — nothing to process. Nice work."

If items exist, also call get_brain_overview to get the list of existing projects for refile suggestions.

Step 2 — Group and Present

Before processing item-by-item, scan all items and group related ones:

  • Items that mention the same project or topic
  • Items captured close together that form a sequence

Present a summary:

"You have [N] items in your inbox. I see [X] that look related to [project], [Y] that might be new topics. Let's process them."

Step 3 — Process Each Item (or Group)

For each item or group of related items:

  1. Show the item with its ID and content
  2. Suggest an action based on content analysis:
    • Refile to [project] — if it clearly matches an existing project
    • Create new project — if it represents a distinct new effort
    • Convert to note — if it's reference material, not an action
    • Add metadata — if it's a task but needs priority/due date
    • Discard — if it's stale, duplicate, or no longer relevant
  3. Wait for confirmation — never refile without the user agreeing

For batches of related items going to the same project, suggest a batch refile:

"These 4 items all look like they belong in [project]. Refile all of them?"

Step 4 — Execute Actions

Based on user decisions:

  • refile_item — for moves to existing projects (use batch array for multiple items)
  • create_project then refile_item — for new project creation
  • create_project_note — for converting items to notes
  • update_todo — for adding priority or due dates after refile
  • delete_todo — for discarding items (confirm first)

Step 5 — Summary

After processing all items:

"Done! Processed [N] items: [X] refiled, [Y] new projects created, [Z] discarded. Inbox is at [remaining] items."

If inbox is now empty, celebrate briefly.

Tool Sequence

get_dump_items()
get_brain_overview()
  → group and present
  → for each item/group:
    → suggest action
    → await confirmation
    → refile_item() / create_project() / create_project_note() / delete_todo()
  → present summary

Notes

  • Speed over perfection — if the user says "yeah, refile all of those", do it in one batch call
  • When suggesting a project for refile, explain why you think it matches ("this mentions the API, which relates to your backend-v2 project")
  • If an item is ambiguous, offer 2-3 options rather than guessing
  • Keep momentum — do not over-discuss each item. Quick decisions are the goal
  • If the inbox has > 15 items, offer to do a "quick pass" (obvious refiles first) vs. "thorough pass" (every item)
  • Tone: systematic and efficient — like processing email with a good filter

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.