What is a Skill?

Getting Started

Create a skill

Just tell your AI assistant what you want. For example:

"Create a skill called code-review that triggers on PR and review.
It should remind me to check for security issues, missing tests,
error handling, and performance problems."

That's it. From now on, whenever you mention "review" or "PR" in any project, the AI automatically follows your checklist. You don't need to remember any syntax — the AI handles the details.

Import skills you already have

Already have a CLAUDE.md, .cursorrules, or any rules file? Just ask:

"Import my skills from CLAUDE.md"
"Import the .cursorrules from this project into my skills locker"

Each section of your file becomes a separate skill. ContextStream figures out the format automatically and skips duplicates if you import the same file again.

Browse and use your skills

"Show me my skills"
"List team skills"
"Run the deploy-checker skill"
"Run standup prep"

How Skills Activate

Skills work in two ways, and you don't have to think about which one — it just happens.

Personal vs Team Skills

Every skill is either yours or your team's. You choose when you create it, and you can change it later.

Want to share a personal skill with your team? Just say "Share my deploy-checker skill with the team."

Skills That Do Things

Most skills are just instructions — text the AI follows. But skills can also do things: fetch your tasks, pull decisions from memory, call APIs, and chain steps together. You describe what the skill should do in plain language when creating it, and ContextStream wires up the actions.

"When this skill runs, also remind me to check the changelog"

"Create a skill that pulls all security-related decisions
and then checks them against the current code"

"Make a standup skill that fetches my in-progress tasks,
completed tasks, and blocked tasks, then summarizes them
in Yesterday / Today / Blockers format"

"Run the deploy checker for staging"
"Run the deploy checker for production"  ← same skill, different behavior

You can also preview what a skill would do without actually running it — just ask for a "dry run."

Bring Your Existing Rules

ContextStream imports from all the common AI tool formats. You don't need to rewrite anything.

Examples

Here are skills real teams use. You can create any of these just by describing what you want.

"Create a team skill called api-standards that activates when we talk
about endpoints or APIs. It should enforce: responses use
{success, data, error, metadata} format, UUIDs for all IDs,
pagination on list endpoints, validation at handler level,
and never expose internal errors to clients."

"Create a skill for pull requests that enforces: titles under 72 chars
with feat:/fix:/refactor: prefix, body must have Summary, Changes,
Testing checklist, and Rollback plan. Always review all commits."

"Create a standup skill that pulls my current tasks and formats them
as Yesterday / Today / Blockers. It should trigger on 'standup'
or 'daily' or 'status update'."

"Create a high-priority team skill for security reviews. When triggered,
it should pull our past security decisions from memory and then check
the current code against OWASP Top 10, input validation, auth on every
endpoint, no hardcoded secrets, and proper CORS config."

History & Tracking

Skills keep track of changes and usage automatically.

Web Dashboard

You can manage skills from the ContextStream dashboard too — no AI assistant required.

• Browse — Search, filter by scope or status, see all your skills at a glance
• Edit — Update instructions, triggers, priority, and actions with a visual editor
• Import — Upload a file or paste content, pick the format, and import
• History — See version history and execution logs for any skill

Access: Sidebar → Skills, or press Cmd+K and type S

Common Questions