Skip to content
Last updated

CLI Agent Power Tools overview

CLI Agent Power Tools for Smartsheet are local sub-agents that use the Smartsheet MCP Server to automate common project-management tasks. Each agent is a markdown file that lives on your machine under .claude/agents/ (Claude Code) or .gemini/agents/ (Gemini CLI).

The following table describes each power tool, noting the kind of work it does, an example prompt, and the manual steps that it automates.

Power ToolStageThe promptWhat it typically replaces
bottleneck-scannerREAD"Who's the bottleneck across my active projects?"Opening sheets, counting tasks by owner, cross-checking thread context
reassignment-helperWRITE"Reassign everything from Alex to Jordan."Filtering each sheet, clicking each row, checking access
engagement-clonerCREATE"Clone this project sheet for a new engagement."Finding a template, copying, clearing data, renaming, re-sharing
risk-scannerSCAN"What's at risk across my active projects?"Manually checking due dates, blocked rows, and unassigned tasks
standup-prepBRIEF"Prep my standup."Pulling up sheets before a meeting to remember what to say
status-comms-writerDRAFT"Draft a Slack update from this risk scan."Translating project data into a message someone else will actually read

These agents differ from Smartsheet's platform sub-agents (Risk Analysis, Resource Optimizer, and others). Those run on Smartsheet's infrastructure and use proprietary execution data. CLI Agent Power Tools run in your local CLI session and call only the public MCP tools your account already has access to.

Local, open-source sub-agents provide the following benefits:

  • Control and transparency. Each agent is a markdown file you can read, edit, and version-control. You know exactly what it does.

  • Composability. Tools chain together naturally — risk-scanner feeds standup-prep feeds status-comms-writer. You control the pipeline.

  • Custom domain logic. Edit any agent file to bake in your team's naming conventions, risk thresholds, or workflow rules.

Important: CLI Agent Power Tools for Smartsheet is an open-source project built and maintained by the Smartsheet AI Platform team. It is provided as-is under an MIT license (refer to the License) and is not an officially supported Smartsheet product.

How do the tools work together?

These tools fall into two natural sequences.

READ → WRITE → CREATE is a trust-building progression. Start by querying your projects for bottlenecks — it's read-only, no risk. Act on what you find by reassigning tasks — preview before committing. Then carry that understanding into something new: a cloned engagement sheet, ready to go. Each step adds consequence, so you build confidence before committing to data changes or creating engagement templates.

SCAN → BRIEF → DRAFT is a daily rhythm. Surface risks across your projects. Carry what's worth raising into your standup. Turn standup content into a status message for other people to read. Each stage filters the signal, so the next stage focuses on priorities.

How do I start using the CLI Agent Power Tools?

See the CLI Agent Power Tools GitHub repository for full setup instructions: installing Claude Code or Gemini CLI, configuring the Smartsheet MCP connection, and cloning the repo.

Once set up, use a prompt like the one below to verify everything is working.

Example: Assessing risks across projects

Prompt:

What's at risk across my active projects in workspace "ABC Projects"?

Response:

(The following responses is an example. Responses can vary.)

  🔴 HIGH RISK (3 signals)

  ┌─────────────────────────────┬────────────────────┬────────────┬──────────────┐
  │            Task             │      Signals       │    Due     │ Owner        │
  ├─────────────────────────────┼────────────────────┼────────────┼──────────────┤
  │ Safety Measures (Site Plan) │ Overdue + Blocked  │ 2026-03-20 │ Aarav Sharma │
  └─────────────────────────────┴────────────────────┴────────────┴──────────────┘

  ---

  🟠 AT RISK (2 signals)

  ┌──────────────────────────────────────────────────┬────────────────────┬────────────┬─────────────────┐
  │                       Task                       │      Signals       │    Due     │      Owner      │
  ├──────────────────────────────────────────────────┼────────────────────┼────────────┼─────────────────┤
  │ Environmental Considerations (Site Plan)         │ Overdue + Blocked  │ 2026-07-20 │ Lukas Hoffmann  │
  ├──────────────────────────────────────────────────┼────────────────────┼────────────┼─────────────────┤
  │ Retaining Structures (Site Plan)                 │ Overdue            │ 2026-07-22 │ Ananya Patel    │
  ├──────────────────────────────────────────────────┼────────────────────┼────────────┼─────────────────┤
  │ Natural materials (Product Presentation)         │ Overdue + No owner │ 2026-08-21 │ —               │
  ├──────────────────────────────────────────────────┼────────────────────┼────────────┼─────────────────┤
  │ Gear collections 5.4 (Product Presentation)      │ Overdue + No owner │ 2026-09-04 │ —               │
  ├──────────────────────────────────────────────────┼────────────────────┼────────────┼─────────────────┤
  │ Environmental display 5.6 (Product Presentation) │ Blocked + No owner │ 2026-10-02 │ —               │
  └──────────────────────────────────────────────────┴────────────────────┴────────────┴─────────────────┘

  ---

  🟡 WATCH (1 signal) — 15 items

  Mostly schedule-only (overdue) tasks spread across Site Plan, In-Store Experience, and Product Presentation. Notable ones:

  - Regulatory Compliance — In Review since 2026-03-20 (Sarah Miller)
  - POS Experience Design — In Review since 2026-05-08 (Wei Chen)
  - Gear collections 5.5 — Not Started, due 2026-09-18, 4 days ago (Elena Rodriguez)
  - Groundbreaking ceremony — No owner, no due date, Not Started

  ---

  Signal Totals

  - Overdue: 19 items
  - Blocked: 4 items
  - Owner gaps: 3 items

  ---

  Want me to hand this to standup-prep to build a pre-meeting brief, or to status-comms-writer to draft a Slack/email update?

The offer to pass the risk assessment to the other power tools reflects how the tools are designed to work: each agent passes its output as context to the next. Accepting the offer hands the full risk scan to standup-prep or status-comms-writer without re-running the query or restating the data.