Skip to content
Back to skills

Walnut Suggest Cards

ASecurity

Emit clickable suggest-action cards in the Walnut web console — buttons wired to real Walnut ops (task triage, focus tier moves, scheduling). Load this BEFORE proposing any one-click action the user should approve rather than type out, e.g. "put this task in Focus?", a daily plan draft, or a batch cleanup with per-item buttons.

  • 32 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 1, 2026
ai-agents

Works with

  • cli

Security analysis

A100/100

Scanned October 1, 2026

npx -y skills add EvanZhang008/open-walnut --skill walnut-suggest-cards --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Walnut Suggest Cards?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Walnut Suggest Cards
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/evanzhang008-walnut-suggest-cards/badge)](https://www.skillsdirectory.com/skills/evanzhang008-walnut-suggest-cards)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: walnut-suggest-cards
description: Emit clickable suggest-action cards in the Walnut web console — buttons wired to real Walnut ops (task triage, focus tier moves, scheduling). Load this BEFORE proposing any one-click action the user should approve rather than type out, e.g. "put this task in Focus?", a daily plan draft, or a batch cleanup with per-item buttons.
---

# Suggest-action cards

The Walnut web console renders `<suggest>` blocks in assistant messages as cards with clickable buttons. A click executes the named op with the given args, under the user's authorization. Every other surface (phone v1 projection, plain-text notifications) shows the raw tags, so the surrounding prose must read fine on its own.

## Syntax

```
<suggest title="Triage this" multi? sticky?>
Markdown body (optional): context, a numbered draft, task refs.
<action tool="task_focus_tier_set" args='{"id":"t_1","tier":"focus"}' label="Put to Focus" style="primary"/>
<action tool="task_update" args='{"id":"t_1","start_date":"2026-09-01"}' label="Push to Sep 1"/>
<action tool="task_delete" args='{"id":"t_1"}' confirm="Cannot be undone" label="Delete" style="danger"/>
<action dismiss label="Ignore"/>
</suggest>
```

## Attributes

- `tool`: any Walnut op name (`walnut tools list`). The click runs it exactly as written — the server validates args against the op's schema and refuses unknown ops.
- `args`: complete, static, valid JSON. REAL ids only — never a placeholder, never a guess.
- `label`: short button text. `style`: `primary`, `danger`, or omitted.
- `confirm="..."`: adds an inline confirmation step. Required for anything irreversible (delete, merge, anything that runs code or replaces a document — the server refuses destructive ops without it).
- `<action dismiss label="..."/>`: closes the card, runs nothing. Include one on every card.
- Card-level `multi`: each action is independently clickable once (batch decisions). `sticky`: the card stays usable after clicks. Default: one card = one decision, the first click settles it.

## Rules

- A card is for a CHOICE the user should approve. Do the work yourself when no approval is needed; never emit a card instead of an answer.
- One decision per card; at most one card per answer unless the user asked for a batch review.
- The body prose must carry the full meaning without the buttons (other surfaces show raw text).

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.