Expanded Systems Reference
Private Skills
Approved operational packages for trusted agent work.
Summary
Where this system fits.
Secondary reference material for teams and crawlers that need more detail than the public navigation exposes.
What It Is
Private Skills package customer-specific instructions, scripts, workflow templates, and approved access paths so trusted agents can perform useful work safely.
What It's For
Approved operational packages for trusted agent work.
Why It's Needed
AI tools become risky when every operator improvises prompts, credentials, scripts, and workflow steps for private business processes.
Reference Detail
Operational shape.
Who Uses It
- Approved operators
- Forward-deployed engineers
- Local agent workspace users
- Teams running repeatable private workflows
Inputs
- Customer-specific operating instructions
- Workflow templates
- Approved scripts and CLI tools
- Source rules
- Permissioned Secure CDM APIs
Processing
- Packages repeatable workflow behavior
- Encodes source rules and operating instructions
- Uses approved APIs instead of ad hoc private-data handling
- Supports customer-installable or Big Robot-managed delivery modes
Human Controls
- Skill approval
- Version checks
- Operator review of outputs
- Permissioned access to workflow APIs
Security and Privacy Notes
- Designed for approved private workspaces and controlled API access
- Avoids embedding secrets or private customer data in public skill content
- Can run locally or in a managed environment depending on customer security needs
Outputs
- Repeatable workflow actions
- Generated reports or artifacts
- Structured updates through approved APIs
- Operator-readable instructions and checks
Integrations
- Secure CDM
- Agent Manager
- Codex-style approved workspaces
- Customer-approved tools and systems
Implementation Pattern
- Identify a repeatable workflow with clear source rules.
- Package the workflow instructions and approved tool paths.
- Add version and access checks.
- Run with human review until the workflow is trusted.
Results or Expected Outcomes
- Less one-off prompting for sensitive workflows
- More consistent agent outputs
- Cleaner operational handoffs between people and agents
When to Use
- A team repeats a private operational workflow
- Source rules and approval boundaries matter
- Agents need a reliable package of instructions and tools
When Not to Use
- The task is not repeatable
- The workflow cannot be bounded with safe access rules
- A public generic assistant is enough
Related Systems
FAQ
FAQ
Are Private Skills public prompts?
No. They are approved operational packages for a customer's private workflow environment, not generic public prompt templates.
Can Big Robot manage Private Skills for a client?
Yes. Private Skills can be customer-installable or operated by Big Robot inside an approved managed environment.