Skip to content

Using the SkillMeat CLI Skill in Agent Workflows

The skillmeat-cli custom Claude Code skill enables agents and automated workflows to manage your SkillMeat artifacts directly within a session. Instead of pausing to run manual CLI commands, agents can discover, deploy, and manage artifacts as part of their task execution.

This guide is distinct from the CLI Commands Reference — it focuses on agent integration and activation rather than command syntax.

Skill vs. CLI Reference

This guide teaches you how to activate and use the skill in an agent workflow. For detailed command syntax, options, and exit codes, see the CLI Commands Reference and CLI Reference Guide.

Overview

What is the skillmeat-cli Skill?

A custom Claude Code skill is a bundle of knowledge and operational patterns that enhances an agent's capabilities. The skillmeat-cli skill allows agents to:

  • Discover and search artifacts across your collection and marketplace
  • Deploy artifacts to projects programmatically
  • Manage your collection — list, inspect, sync, and update artifacts
  • Create and manage bundles — group and publish artifact sets
  • Scaffold projects from templates and context
  • Capture and consume memories for learning and context preservation
  • Handle authentication — OAuth login, PAT storage, and credential lifecycle
  • Support enterprise workflows — migration, BOM signing, attestation

Why Use the Skill?

Without the skill: You must manually run SkillMeat CLI commands between tasks.

# Manual workflow — interrupts agent progress
skillmeat search python-testing
skillmeat deploy pytest-plugin --to /project

With the skill: Agents orchestrate SkillMeat operations seamlessly.

Agent Task: "Set up testing for this Python project"
↓
Agent invokes skillmeat-cli skill
↓
Agent discovers pytest plugins, deploys them, captures the memory
↓
Task complete, no manual CLI invocation needed

Prerequisites

Before using the skillmeat-cli skill, ensure:

  1. Claude Code CLI installed — See Claude Code documentation for installation
  2. SkillMeat CLI installed — See Quickstart Guide for installation
  3. SkillMeat collection initialized — Run skillmeat init (see Collection Init & Configuration)
  4. Skill deployed to your project or user scope — See "Activate the Skill" section below

Verify Installation

Check that SkillMeat is installed and accessible: 

skillmeat --version

Activate the Skill in a Session

Option 1: Invoke Explicitly in a Task Prompt

Include a Skill() call in your task to agents:

Task("python-backend-engineer", "Skill("skillmeat-cli")
Implement a new API endpoint for user authentication.
Consider reusing the existing auth-flow agent from the marketplace.")

The agent loads the skill and gains access to all artifact management patterns.

Option 2: Via Claude Code Slash Command

In Claude Code, trigger the skill with:

/skill:skillmeat-cli

Then provide your task to the agent.

Option 3: Reference in Implementation Plans

When using .claude/progress/ or .claude/plans/ files, include the skill in your implementation plan:

tasks:
  - id: FEATURE-1.1
    title: Set up testing infrastructure
    assigned_to: python-backend-engineer
    skills_required:
      - skillmeat-cli
    ...

Example Agent Workflows

Workflow 1: Agent Discovers and Deploys a Needed Artifact Mid-Task

Scenario: A backend engineer is implementing a new service but needs a utility skill they haven't yet added to the project.

Agent: "I need to integrate credential validation. Let me search for an existing credential-checker skill..."

[Agent invokes skillmeat-cli skill]

Agent: Discovers anthropics/skills/credential-validation in the marketplace
Agent: "Found credential-validation. Deploying to ./.claude/skills/..."
Agent: Deploys the skill locally
Agent: Resumes implementation, now using the deployed skill
Agent: Captures a memory: "credential-validation skill deployed; pattern used in /services/auth.py lines 45-62"

Agent prompt: 

Skill("skillmeat-cli")

Implement user credential validation in the auth service.
First, search for and deploy any existing credential-checking skills or patterns
from the marketplace. Then implement the validation logic using those patterns.
Capture a memory of what you deployed and where you used it.

Workflow 2: Agent Captures a Memory After a Bug Fix

Scenario: An agent discovers and fixes a tricky bug, then shares that learning for future agents.

Agent: "Bug found: ListItemCreate requires list_id in the request body, not as a query param."
Agent: "Fixing the schema in skillmeat/api/schemas/list_item.py..."
Agent: "Invoking skillmeat memory command to capture this gotcha..."

[Agent runs: skillmeat memory item create ...]

Agent: "Memory captured for future agents: 'ListItemCreate list_id must come from URL path, not request body'"

Agent prompt: 

Skill("skillmeat-cli")

Debug why the API returns a 422 error when creating list items.
Once fixed, capture a memory with type "gotcha" so future agents avoid this trap.
Include the file path and line numbers in the memory anchor.

Workflow 3: Agent Searches Collection for Prior Patterns Before Implementation

Scenario: A frontend engineer searches the collection for component patterns before building a new UI.

Agent: "Let me search the collection for existing data table components..."

[Agent invokes skillmeat-cli skill]

Agent: Discovers @miethe/ui DataTable component
Agent: "Found DataTable with sorting, filtering, and pagination built-in."
Agent: "Deploying to the project and reviewing the pattern..."
Agent: Implements the new feature using the existing DataTable pattern

Agent prompt: 

Skill("skillmeat-cli")

Implement a filterable, paginated artifact list for the dashboard.
First, search the collection for existing table/list components.
Then implement using those patterns to ensure consistency.

Key Commands for Agents

These are the most common skillmeat subcommands agents use. For full syntax and options, see CLI Commands Reference.

Command When to Use Example
skillmeat search <query> Find artifacts by keyword or semantic match skillmeat search python-testing
skillmeat list Show all artifacts in the collection skillmeat list
skillmeat show <artifact> Inspect a specific artifact's metadata skillmeat show canvas
skillmeat deploy <artifact> --to <path> Deploy to a project directory skillmeat deploy canvas --to .
skillmeat add <source> Add an artifact from GitHub or local path skillmeat add skill anthropics/skills/canvas
skillmeat memory item create Capture a learning (with API fallback) skillmeat memory item create --type learning --content "..."
skillmeat memory search <query> Find prior captured memories skillmeat memory search "auth pattern"
skillmeat scaffold --from-bundle Render project files from a bundle skillmeat scaffold my-bundle --output .
skillmeat bundle create Create a new artifact bundle skillmeat bundle create --name my-bundle
skillmeat history View activity history for an artifact skillmeat history canvas
skillmeat snapshot Create a backup of your collection skillmeat snapshot "Before refactor"

Command Syntax Details

The examples above show typical usage. For all flags, arguments, and options, see CLI Commands Reference and CLI Reference.

Scope Selection: User vs. Local

When deploying or adding artifacts, agents must choose a scope:

Scope Location Use When
user (default) ~/.skillmeat/collection/ Artifact should be available in all projects globally
local ./.claude/ in current project Artifact is specific to one project only

Example:

# Deploy to global user scope (all projects)
skillmeat deploy canvas --scope user

# Deploy to local project scope
skillmeat deploy canvas --scope local --to /path/to/project

When agents are working on a specific project, they typically use --scope local --to . to avoid polluting the global collection.

Authoring Your Own Agent Integrations

If you're building custom agents that should integrate with SkillMeat:

  1. Load the skill explicitly: 

    Skill("skillmeat-cli")

  2. Route the agent to the relevant workflow based on intent:

  3. Discovering artifacts → use discovery patterns
  4. Deploying → use deployment patterns
  5. Managing collection → use management patterns

  6. Capturing memories → use memory patterns

  7. Reference canonical docs for command syntax:

  8. Always cite CLI Commands Reference or CLI Reference when providing command examples

  9. Do not invent command syntax — agents should verify against the canonical reference

  10. Handle the memory item create fallback:

  11. If skillmeat memory item create returns a 422 or 400 error, the agent should automatically fall back to the API endpoint documented in the Memory System guide

  12. Capture learnings opportunistically:

  13. Whenever an agent discovers a pattern, bug fix, or architectural decision, encourage it to capture that as a memory for future agents
  14. Include file anchors so memories are linked to specific code locations

Next Steps

Troubleshooting

Skill Not Found or Not Loaded

Problem: Agent reports "skillmeat-cli skill not found"

Solution: 1. Verify the skill is deployed: skillmeat list --scope user 2. If missing, deploy it manually: skillmeat deploy skillmeat-cli --scope user 3. Re-run your agent task

Memory Item Create Returns 422

Problem: skillmeat memory item create fails with 422/400 error

Solution: The CLI has a known argument parsing issue. The agent should automatically fall back to the API endpoint:

curl -s "http://localhost:8080/api/v1/memory-items?project_id=<BASE64_PROJECT_ID>" \
  -X POST -H "Content-Type: application/json" -d '{
  "type": "gotcha",
  "content": "Your learning here",
  "confidence": 0.85,
  "status": "candidate",
  "anchors": ["path/to/file:code:10-30"]
}'

For details, see the Memory System guide.

Rate Limit Errors

Problem: skillmeat commands hit GitHub rate limits

Solution: Configure a GitHub personal access token (see Collection Init & Configuration):

skillmeat config set github-token ghp_your_token_here

This increases your rate limit from 60 to 5,000 requests per hour.

FAQ

Q: Do I need to use the skill, or can I just run CLI commands manually?

A: Both work. Manual CLI invocation is fine for one-off commands. Use the skill when you want agents to orchestrate artifact management within a task, without manual interruption.

Q: Can multiple agents use the skill in the same session?

A: Yes. Each agent can independently invoke Skill("skillmeat-cli") and orchestrate operations. They will share your collection and respect the same scopes (user vs. local).

Q: Does the skill require authentication?

A: SkillMeat runs in zero-auth local mode by default. For team or production setups, see Authentication Setup.

Q: What if an agent's skillmeat command fails?

A: The agent should capture the error and attempt recovery. For systematic issues (rate limits, network failures), see CLI Reference — Exit Codes.