Claude Integration

Built-in Skills

The five slash commands that give Claude direct control over tasks, projects, memory, system health, and skill creation.

What Are Skills?

Skills are slash commands that Claude can run during a conversation. Each skill is either a Python script that calls the Cognova API or a pure instruction file that guides Claude's behavior. When you type /task create "Fix login bug", Claude executes the task skill's Python script, which sends a POST request to the Cognova API.

Skills live in ~/.claude/skills/ and are registered automatically by Claude Code.

/task -- Task Management

Create, list, update, complete, and delete tasks.

Subcommands

CommandDescription
create <title>Create a new task
listList tasks with optional filters
update <id>Update task fields
done <id_or_search>Mark a task as complete
delete <id>Soft-delete a task

Flags

create:

  • --project, -p -- Associate with a project (by name, fuzzy matched)
  • --priority, -P -- Priority level: 1 (low), 2 (medium), 3 (high)
  • --due, -d -- Due date (today, tomorrow, next week, day name, or YYYY-MM-DD)
  • --tags, -t -- Comma-separated tags
  • --description -- Longer description text

list:

  • --status, -s -- Filter by todo, in_progress, done, blocked
  • --project, -p -- Filter by project name
  • --search, -q -- Search title and description
  • --due -- Filter by today, week, or overdue

update: Accepts --title, --description, --status, --priority, --due, --tags, --project

Examples

/task create "Deploy staging environment" --project homelab --priority 3 --due friday
/task list --status todo --due week
/task done "staging"

The done command accepts either a task ID prefix or a title search. If the search matches a single incomplete task, it completes it. If multiple match, it shows options.

Task IDs use UUID prefixes. You only need the first few characters to identify a task (e.g., a3df instead of the full UUID).

/project -- Project Management

Organize tasks into projects with names and colors.

Subcommands

CommandDescription
search <query>Find projects by name
listList all projects
create <name> --color <hex>Create a new project
update <id>Update project fields
delete <id>Soft-delete a project

Flags

create:

  • --color, -c (required) -- Hex color code (e.g., #3b82f6)
  • --description, -d -- Project description

update: Accepts --name, --color, --description

Examples

/project create "Homelab" --color #10b981 --description "Self-hosted infrastructure"
/project search home
/project list

The create command checks for duplicate names before creating. If a project with a similar name exists, it warns you and suggests using the existing one.

/memory -- Memory Management

Search, store, and inspect the persistent memory system.

Subcommands

CommandDescription
search <query>Search memories by content
recent [limit]Get the most recent memories
store <content>Manually store a memory
decisionsList all decision-type memories
about <topic>Find everything known about a topic
contextPreview what context would be injected into a session

Flags

search / recent / decisions:

  • --type, -t -- Filter by decision, fact, solution, pattern, preference, summary
  • --project, -p -- Filter by project path
  • --limit, -l -- Max results (default: 10-20)

store:

  • --type, -t -- Memory type (default: fact)
  • --project, -p -- Associated project path
  • --relevance, -r -- Relevance score 0-1 (default: 0.9)

Examples

/memory about "deployment"
/memory store --type decision "Using Caddy instead of Nginx for reverse proxy"
/memory search "database" --type solution
/memory decisions --project homelab

Claude is instructed to use /memory proactively. Before starting work, it checks what it already knows. After completing work, it stores outcomes and decisions. See Memory System for how this works under the hood.

/environment -- System Health

Inspect the Cognova installation, check service status, and troubleshoot issues.

Subcommands

CommandDescription
infoShow system details (OS, Node, Python, disk, vault stats)
statusCheck PM2 process status, memory usage, API health
logs [--lines N]Show recent PM2 log output (default: 30 lines)
healthRun a full API health check across all endpoints

Examples

/environment status
/environment health
/environment logs --lines 50

The health subcommand tests five API endpoints (health, tasks, projects, documents, memory) and reports response times for each. This is the fastest way to diagnose whether the Cognova server is running correctly.

=== API Health Check ===

Base URL: http://localhost:3000

  OK    Health            12ms
  OK    Tasks             45ms
  OK    Projects          38ms
  OK    Documents         52ms
  OK    Memory            41ms

Overall: All healthy

/skill-creator -- Build New Skills

An interactive wizard that helps you design and create custom Claude Code skills.

Unlike the other skills, /skill-creator is a pure instruction skill -- it has no Python script. Instead, it provides Claude with a structured process for gathering requirements, checking for existing solutions, and generating the right skill files.

What It Does

  1. Gathers requirements -- asks what the skill should do, who invokes it, and whether it needs external APIs
  2. Searches for existing solutions -- checks for MCPs, plugins, and community tools before building from scratch
  3. Determines skill type -- pure instruction, Python script, or subagent
  4. Generates the files -- creates SKILL.md and optionally a Python script with proper structure
  5. Handles secrets -- shows how to use the Cognova secrets store instead of hardcoding API keys

Example

You: /skill-creator
Claude: What kind of skill would you like to create?
You: I want a skill that sends a summary of my completed tasks to Discord every week
Claude: Let me check if there's an existing MCP for Discord webhooks...
       [searches, finds options, suggests approach]
       I'll create a Python script skill that uses the Cognova API
       for tasks and a Discord webhook for delivery. Let me set that up.

See Custom Skills for the full guide on skill development.