Bitsight Skills provides downloadable skills that help AI agents work with Bitsight data and APIs. A skill gives an agent instructions for a specific task so you can ask for the outcome in natural language.
You can browse the available skills in the Skills Library or access a subset in the Bitsight skills GitHub repository.
What are agent skills?
Agent skills are reusable instructions that help an AI assistant complete a particular task. Each skill includes a .SKILL.md file that tells the agent what to do. A skill may also include scripts, templates, or other supporting files. When a request is relevant to an installed skill, the agent can select it automatically.
Why use agent skills?
Skills give an agent the context and steps it needs to carry out a task consistently. They can provide,
- Domain expertise: Instructions and resources for a specific subject.
- Repeatable workflows: A consistent process for tasks with several steps.
- Reuse across agents: A skill can be used with agents that support the skill format.
How do I install an agent skill?
- Open the Bitsight Skills Library and download the skill ZIP file.
- Extract or upload the ZIP file.
- Copy the extracted skill folder into the skills directory for your AI agent. For example:
- Claude Code: ~/.claude/skills/ for your user or .claude/skills/ for a project.
- OpenAI Codex CLI: ~/.codex/skills/.
Your agent discovers skills installed in its supported directory. Skill installation tools may also automate this process.
What do I need to use Bitsight skills?
Requirements depend on the skill:
- For an API-based skill, you need a Bitsight API token. To get or generate your user token, go to Account → User Preferences → User API Token. For token instructions, see API Token Management.
- Skills tagged mcp-friendly support the Bitsight MCP. To use one, you need a Bitsight MCP connection with access to the data and tools the skill requires. Check the skill's instructions for its specific setup requirements.
Do Bitsight skills use the MCP or the API?
At this time, some Bitsight skills support the Bitsight MCP, while others work directly with Bitsight APIs only. A skill's .SKILL.md explains which connection it uses and what you need to set up before running it. Skills tagged mcp-friendly support the Bitsight MCP. Downloading and adding a skill does not set up its MCP connection or API authentication for you.
How do agent skills work?
Agents load skills in three stages:
- Discovery: The agent reads each skill's name and description to identify relevant skills.
-
Activation: When a task matches a skill, the agent loads its full
.SKILL.mdinstructions. - Execution: The agent follows those instructions, which may include using bundled scripts or supporting files.
Loading the full instructions only when needed lets an agent have access to many skills without loading all of them at once.
Can I use multiple skills together?
Yes. Skills are modular, so an agent can select and coordinate more than one skill to complete a larger workflow.
Can I create my own skills?
Yes. You can create custom skills that use Bitsight APIs. A skill needs a .SKILL.md file with its instructions and can also include scripts or templates. The option to contribute your own skills to the Bitsight Skills Library is not currently available.
How are skills different from slash commands?
A skill can be selected automatically based on your request. In Claude Code, you can also invoke an installed, user-invocable skill directly by typing /<skill-name>. A slash command is an explicit way to invoke a skill or command; some skills may be configured for automatic use only.
How often are Bitsight skills updated?
The Bitsight team updates the Skills Library regularly. Check the Last updated date on a skill's card to see when that skill was most recently updated.
Are all available skills in the GitHub repository, or are some only available in the platform?
The public GitHub repository contains a selection of Bitsight skills. Skills that use proprietary Bitsight capabilities are available inside the platform after you log in, so access is limited to authorized users.
- October 5, 2026: Published.
Comments
Please sign in to leave a comment.