Skip to main content

Library

The Library is the centralized management system for prompts, skills, imprints and artifacts in Strongly.AI. Its main sections:

  • Prompts -- System prompts, user prompts, and templates with versioning, variables, and model comparison
  • Skills -- Reusable knowledge packages (SKILL.md + supporting files) that teach agents how to handle specific tasks
  • Imprints -- Named bundles of library items an agent learns in one step
  • Artifacts -- Reports, dashboards, plans and documents your agents save, or that you upload

Prompts​

Prompt Types​

TypeDescriptionUse Case
System PromptDefines AI model behavior and personalityAgent system instructions, chatbot personas
User PromptQueries or instructions sent to modelsExtraction templates, analysis requests
TemplatePre-built prompt structures with variablesSummarization, classification, code generation

The Prompts Page​

Library > Prompts lists every prompt you may use: your own, those shared with you, those open to all users, and the platform prompts (the operating prompts the reference agents use, owned by Platform).

  • Summary cards: Total Prompts, System Prompts, User Prompts and Templates count your whole library, whatever page, filter or search the list shows.
  • Type filter: All, System Prompts, User Prompts, Templates.
  • Tag filter: only prompts with every tag you add.
  • Search: finds prompts by meaning and by words at once, best match first. Type a prompt's name, or describe what it does ("help customers with their problems").
  • Actions: View and Duplicate on every prompt; Edit and Delete on the prompts you may change.

Key Features​

  • Automatic Versioning: Every change to a prompt's content creates a new version with its change note
  • Variable Templating: Use {{variableName}} syntax for reusable, parameterized templates
  • Model Comparison: Test prompts against multiple AI Gateway models side-by-side
  • Sharing & Visibility: Share with specific users or allow all users of your organization

Quick Start​

  1. Navigate to Library > Prompts in the sidebar
  2. Click New Prompt
  3. Select a prompt type from the dropdown
  4. Enter a name and optional description
  5. Write your prompt content in the editor
  6. Give any detected variable a default value (optional)
  7. Add tags for organization (comma-separated)
  8. Click Save Prompt

Variables​

Write a variable as {{name}} anywhere in the content. As you type, the form lists each variable it detects under Detected Variables, where you can give it a default value. On the prompt's page the Variables table shows each one's default and whether it is Required: a variable with no default is required, and rendering the prompt (a workflow node, the REST API) without a value for it is refused, naming it. A variable with a default takes the default when no value is given.

Platform Prompts​

Platform prompts are read only. To change one for your own agents, click Duplicate: the copy is your own private prompt (owned by you, without the platform tag), which you can edit and share.

Skills​

Skills are structured knowledge packages that teach agents how to handle specific tasks. Each skill consists of a main SKILL.md file and optional supporting files (references, scripts, assets).

Skill Structure​

my-skill/
├── SKILL.md          # Required -- main skill instructions
├── references/       # Optional -- documentation, API docs
│   └── api-docs.md
├── scripts/          # Optional -- executable code
│   └── validate.py
└── assets/           # Optional -- templates, config files
    └── template.txt

Key Features​

  • Multi-file Support: Main SKILL.md + supporting references, scripts, and assets
  • GitHub Import: Import skills directly from public GitHub repositories
  • Agent Associations: Assign skills to agents with per-agent edit permissions (editable or locked)
  • Versioning: Every change to the instructions is a version; view any version's content and restore it (Restore is offered where you may edit)
  • Quality Assessment: The Assessment tab scores the skill against Anthropic's skill-building guide, with recommendations. Each check passes, warns (part of its points) or fails; a check that does not apply to the skill (for example a cohesion check on very short instructions) is shown as not evaluated and counts toward nothing. When a ceiling holds the score down (a critical check failing caps it at 50, for example), the tab says which and why
  • Sharing: Share skills across your organization

Creating a Skill​

  1. Navigate to Library > Skills in the sidebar
  2. Click Create Skill or Import from GitHub
  3. Enter a name and description. The name is kebab-case (lowercase letters, digits and hyphens, starting with a letter, up to 64 characters, for example unit-converter): agents find the skill by it, and any other name is refused
  4. Write the instructions in markdown (the name and description become the SKILL.md frontmatter; a pasted frontmatter is dropped)
  5. Optionally add supporting files
  6. Click Save

Importing from GitHub​

  1. Navigate to Library > Skills > Import from GitHub
  2. Paste a public GitHub URL (repo root, subdirectory, or specific branch)
  3. The importer reads the folder's SKILL.md and every text file under it as supporting files; a folder without SKILL.md is refused, and binary files (images, fonts) are listed as not imported. The skill is named by the name in its frontmatter (the folder's name when it has none), which must be kebab-case
  4. Click Import Skill. If the same skill is already in your Library, the import says so and leaves it unchanged

Agent-Skill Associations​

Skills can be associated with agents in the workflow editor's Skills tab:

  • Editable: Agent can read and update the skill
  • Locked: Agent can read the skill but cannot modify it (its updates, deletes and restores are refused)
  • Auto-connected: When an agent creates a skill, it's automatically associated with edit access

Artifacts​

Artifacts are the reports, dashboards, plans and documents your agents save, and files you upload. Bodies are stored encrypted (SSE-KMS).

  1. Navigate to Library > Artifacts and click Upload artifact
  2. Give it a title, type, optional summary and tags
  3. Under Content, choose a file (up to 100 MB; its MIME type comes from the file), or switch on Paste text to type the body and pick its MIME type
  4. Click Save artifact

An artifact's page shows its Content (open it in a new tab or download it), Permissions and Metadata (type, size, MIME type, owner, tags, which agents it belongs to, and how it was made). Share and Delete are offered on the artifacts you may change; deleting removes every version's stored body.

Imprints​

An Imprint is a named bundle of library items (skills, memories, rules, tasks, prompts, preferences and artifacts) that an agent learns in one step.

  1. Navigate to Library > Imprints and click New Imprint; give it a name and an optional description
  2. On the imprint's page, click Add content, pick a type and click Add beside each of your items to bundle (the × beside an item takes it out again)
  3. Under Install on an agent, install it on any of your agents; restart the agent to apply

Each card on the Imprints page counts the imprint's items of every type. Items in an imprint are read-only to agents: only you can edit them, in the app. Deleting an imprint keeps its items; they lose the grouping, and agents that had it installed stop reading it. Share an imprint from its Imprint Permissions section like any library item.

REST API​

Full programmatic access is available via the REST API: see the Prompts API and the Skills API.

Access the Library from the sidebar dropdown:

📚 Library
   ├── Prompts     → /library/prompts
   └── Skills      → /library/skills