What Are Skynet Projects?
A Skynet Project is a shared container for collaborative work. It groups collective memory, skills, and agents around one workstream, so your team works inside the same context instead of starting from scratch every time.
- Collective memory what the project knows
- Skills what it can do automatically
- Agents who does the work
- Memory feeds the skills
- Skills feed the agents
- Agents add back to memory
Join a project and you inherit the full context. Run an agent inside it and the agent draws on the project’s memory and skills. Projects are available on Teams and Enterprise plans.
Collective Memory Inside Projects
Every project carries its own collective memory, built automatically as your team works. There is no manual data entry, no knowledge base to maintain, no tagging protocol.
Your tools, untouched
- Slack · Teams
- Meetings
- Jira · Linear
- GitHub
- Docs · Drive
- Chat
- Agents
- Mini-apps
Ask what the team decided about the API redesign and you get the decision along with the exact message or page it came from.
Every Project Is Its Own Scope
- Finance Its own memory Finance team
- Design Its own memory Design team
- Board M&A Its own memory Named members only
No shared wall between them. Context does not cross unless you share it.
Project memory is scoped to one workstream, grouping memory, skills, and agents together. Team memory is the broader shared layer across everything a team does. A project is a focused subset.
Controlling What a Project Remembers
- Connect a source Authorize the scope: which channels, folders, or repositories the project may read
- Set what is remembered What Skynet stores persistently, against what it only reads in the moment and discards
- Query in natural language Ask "what did we decide about the pricing change?" and get an answer with its sources cited
- Let it compound Every interaction sharpens the next answer, and corrections carry forward for the whole team
You are choosing what the project is allowed to keep, not just what it is allowed to see.
Inspect
See exactly what the project has remembered, and where each piece of context came from.
Scope
Limit memory to specific contexts, projects, or timeframes. Retention is set per source.
Delete
Remove anything from memory at any time, for everyone who draws on it.
On Teams and Enterprise plans, role-based access control governs who can create skills, access team memory, and configure connectors. Admins define what enters the company brain, and individual work stays private by default.
Skills Within Projects
A skill is a reusable automation built from a real workflow. The Copilot watches how your team works and proposes one when it spots a repeating pattern.
- The Copilot spots a repeating pattern in how your team works
- It proposes turning that pattern into a skill
- You review and approve
- The skill is built into the project
- It can then be triggered automatically or on demand
Skills are shared across the project. One person builds a skill, everyone benefits, and the more you use a project the less manual work there is. They are not marketplace templates, they come from your actual work.
Parallel Sub-Agents Inside Projects
When an agent inside a project hits work that can be split, it forks sub-agents and merges their results.
- Sub-agent inherits the skill chain
- Sub-agent inherits the skill chain
- Sub-agent inherits the skill chain
- An agent receives an objective within the project
- It identifies components of the work that can run in parallel
- It forks sub-agents Each inherits the project's skill and knowledge chain
- Each sub-agent works independently on its assigned piece
- Results are merged back into a single, coherent output
- You review the final result
The skill and knowledge chain is what keeps this coherent. Every sub-agent starts with everything the project knows, plus the specific piece it was assigned, which is why thousands of tasks can run at once without losing quality.
Team Collaboration and Access Control
Projects are built for teams, but not everyone needs access to everything.
Default access
Project access follows your organizational chart. If you are part of a team, you see that team's projects.
Project-level Defense
For sensitive workstreams, restrict visibility to named members only. Even organization admins can be excluded.
Permission-aware
If someone cannot open a document in the source system, they do not get answers derived from it inside the project.
You defend what needs defending. You share what should be shared.
What Enterprises Get
Visibility
See what projects exist, who is working on them, and what agents and skills are active. Governance, not surveillance.
Security boundaries
Restricted projects stay invisible to anyone not explicitly named, which is critical where data segregation is a requirement.
Audits and roadmaps
AI audits and deployment strategy roadmaps, so projects, agents, and skills roll out with a plan.
What Small Teams Get
No enterprise configuration required. Create a project, invite your team, start working. Memory builds automatically, skills accumulate as the Copilot learns your patterns, and agents run with full context.
The key advantage is sharing. One person builds a skill, the whole team uses it. That lets a small team scale productivity without adding headcount.
How to Think About Projects
A project
Where a specific workstream lives.
Collective memory
What the project knows.
Skills
What the project can do automatically.
Agents
Who does the work inside the project.
Access control
Who can see and participate in the project.
Everything is connected. The memory feeds the skills, the skills feed the agents, and the agents produce work that adds back to the memory. The longer your team works inside a project, the more capable it becomes.