The create skill method symbol. This is the fully qualified name of the method that can be used to create a skill. It must be convertible to a std::function>()>. The method must be declared in the header file for a skill library provided to the cc_skill() build rule as the skill dependency.

This method is used at build time to generate the skill service main.

For the NoOpSkill example with:

namespace intrinsic::skills { class NoOpSkill { public: static std::unique_ptr CreateSkill(); // ... }; } // namespace intrinsic::skills

The Create() method should be registered in the manifest.textproto as:

create_skill: "::intrinsic::skills::NoOpSkill::CreateSkill"

The generated skill service will create skills by invoking this method.

Required interfaces that Services must satisfy for the Skill.

Skill-internal data from a previous call to Projector.Predict.

Logging context of the operation (e.g. IDs of related plan instance, deployment configs etc.).

ID of the world in the world service which holds the initial world state for the skill execution.

Contains optional returned data for the Skill. If a Skill returns data in this field, there will also be descriptor information for the type of data in the Skill information retrieved from the registry.

True if the skill supports cooperative cancellation during execution.

The OpenCensus Trace ID associated with this skill execution. A 32 character hex string.

Error information if the request failed for any reason. error_code and error_message are not set if extended_status is set.

On error, optionally detailed error information.

Prerequisites for a skill refer to the necessary conditions and requirements that must be met for the skill to function properly. These can include specific hardware components like sensors or controllers, initial calibration and setup procedures, software dependencies, network configurations, required permissions and access and/or relevant configuration files.

A usage example of a skill outlines the specific steps and conditions under which the skill is utilized, showcasing its functionality and effectiveness. If the usage example has lengthy details, the author can provide a link to the documentation that shows a practical demonstration of how the skill can be applied in a real-world scenario.

This field describes the error handling strategy for the skill. It can include information about how to handle specific errors, such as retry strategies or recovery procedures.

One or more volume reservations.

A special value - true indicates that all entities in the entire universe are held, implying that this holder of this footprint cannot execute at the same time as anything else that holds any entity.

NOTE: If set, the volume and entity constraints are ignored and every entity is locked with write access.

One or more reservations in the world (objects or frames).

By default we will automatically add the extra missing permissions to make this footprint a complete set. This includes read access for all entities in the parent chains of any already specified entities. As well as write access to any children of write enabled entitiees already provided.

If set to true, this process will be skipped.

One or more resource reservations.

Error information if the request failed for any reason. error_code and error_message are not set if extended_status is set.

On error, optionally detailed error information.

The identifier for the skill as in intrinsic_proto.skills.Skill.id. This must refer to a parameterizable behavior tree registered via RegisterOrUpdateBehaviorTree.

ID of the world in the world service which holds the initial world state for getting the skill footprint.

Logging context of the operation (e.g. IDs of related plan instance, deployment configs etc.).

Skill-internal data from a previous call to Predict.

The resource the instance will use.

An optional instance id for the skill. If empty, a unique id will be generated by the registry.

The skill's id. This can be either the id or the id_version.

The returned instance. The contents of this field will only be valid when status is OK.

The identifier for the skill in the form <package>.<name>.

This should be something like ai.intrinsic.hello_world

Information about the Skill.

The returned skills. The contents of this field will only be valid when status is OK.

Upper limit for the returned page size, i.e. the service might return fewer results.

Default behavior: returns a page size of 20. A page_size of 0 will result in default behavior.

Maximum value is 200. If this is exceeded, returns 200.

Negative values are invalid and will result in an INVALID_ARGUMENT error.

If the page token is empty, up to page_size first elements of the collection will be returned. On subsequent requests, set the next_page_token obtained from the previous call to retrieve the next page.

A filter expression to filter the returned skills.

The format of the filter expression follows the specification outlined in https://google.aip.dev/160. Only the following field are supported: * sideloaded

Skills matching the request.

Pass this token to the subsequent list requests in to obtain the next page.

The type of reservation lock.

Name of an object.

Name of a frame and its parent object.

Indicates whether the skill supports cooperative cancellation during execution. True if the skill supports cancellation, false otherwise.

Sets the cancellation ready time for the skill during execution.

During execution, if the skill is cancelled, its ExecuteContext waits for at most this timeout duration for the skill to have called Ready() (C++) or ready() (Python) before raising a timeout error.

If this timeout is not specified, then the default used by the skill service is 30 seconds.

Sets the execution timeout for the skill.

If this timeout is not specified, then the default used by the skill service is 180 seconds.

Specifies the default values used by the Skill

The set of FileDescriptors needed to dynamically create a message from a factory

New users of the Skill APIs should prefer to retrieve comments from the parameter_descriptor_fileset instead of from the parameter_field_comments.

The leading comments of the fields in the parameter message. The key is the full name of the field.

The fully-qualified name of the Protobuf message

The default value used by the skill for this

ID of the world in the world service which holds the initial world state for the prediction.

Logging context of the operation (e.g. IDs of related plan instance, deployment configs etc.).

Optional skill-internal data from a previous call to Predict.

The possible set of outcomes given the skill and its parameters. This list of outcomes should represent a probability distribution (i.e. the probability fields of each Prediction should sum to 1.0). Each predicted outcome contains a sequence of world updates with associated types. These updates should be confined to the resources available in the footprint that would be generated from a call to GetFootprint() with the same arguments.

Skill-internal data to pass to Projector.GetFootprint and Executor.StartExecute.

Probability that this prediction will happen.

The expected return values from the skill. Has the same caveats as the result in intrinsic_proto.skills.ExecuteResult.

The expected world states during the execution of the skill. Updates are ordered such that each subsequent update should be applied to the previous one after the elapsed number of seconds. The first update should be applied to the initial world used for generating the prediction.

Error information if the request failed for any reason. error_code and error_message are not set if extended_status is set.

On error, optionally detailed error information.

Skill-internal data from a previous call to Projector.Predict.

ID of the world in the world service which holds the initial world state for the skill preview.

Logging context of the operation (e.g. IDs of related plan instance, deployment configs etc.).

The expected return values from the skill. Has the same caveats as the result in intrinsic_proto.skills.ExecuteResult.

The expected world states during execution of the skill. Updates are ordered such that each subsequent update should be applied at the specified time and for the specified duration. The first update should be applied to the initial world used for generating the preview.

The OpenCensus Trace ID associated with this skill preview. A 32 character hex string.

Error information if the request failed for any reason. error_code and error_message are not set if extended_status is set.

On error, optionally detailed error information.

A complete descriptor set for the skill's parameter and execute_result proto messages.

A container image for the skill with its service.

A parameterizable behavior tree.

Metadata associated with this skill (e.g., id, documentation, etc.)

Details about how the skill is to be used.

The processed version of the skill assets. This can be used to send a complete description of a skill using a manifest.

Id of the world in the world service which holds the initial world state for the skill projection.

OPTIONAL. Context of the projection of the skill instance (e.g. IDs of related plan instance, deployment configs etc.).

OPTIONAL. Skill-internal data about the outcome of planning the skill. This data is optional and if provided can be used to make planning faster. It is intended to be used if the skill was previously planned and needs replanning.

Skill-internal data about the outcome of planning the skill. Attempting to execute a skill without passing this data may result in behavior that does not match this ProjectResult.

Error information if the request failed for any reason.

Used to construct a unique topic name for publishing. Instances of this skill will publish to topic /skill///.

Optional documentation that can be shown to users of the skill. It may include information about the data published on this topic.

The published proto's full message name.

The module where the skill's create function is declared.

The module where the skill's proto message schemas are declared.

The symbol for the method to create the skill. This is commonly the skill's constructor. For example, for a skill declared in: intrinsic/skills/examples/my_skill.py

class MySkill(skl.Skill): def init(self): pass

this field should be set to: "intrinsic.skills.examples.my_skill.MySkill" to use the class's constructor to create the skill.

The registration information for the BehaviorTree

The registration information for the skill.

The id of the skill to remove.

The type of reservation.

Name of the equipment to be locked.

The selector only matches Services that provide all of the gRPC services specified here.

Each capability name must be the fully qualified name of a gRPC service.

Specifies how the Service will be used.

NOTE that the sharing type is not enforced. It only specifies the Skill's intent.

The fully-qualified name of the Protobuf message

The set of FileDescriptors needed to dynamically create a message from a factory. More precisely, this is the smallest set containing the file descriptor, A, with the message definition for the return value, along with the file descriptors for all transitive dependencies of file descriptor A.

New users of the Skill APIs should prefer to retrieve comments from the descriptor_fileset instead of from the return_value_field_comments.

The leading comments of the fields in the return value message. The key is the full name of the field.

A well-known name for this skill.

Maps resource slot name to a selector describing handles that can satisfy the implementation's requirements for that slot.

Description of the skill.

Description information for the parameters of this Skill.

Description information for the data returned by a Skill's Execution. This is only present if a Skill utilizes return values.

The package name for the skill. This is dot-separated string describing the package for the skill.

The skill id, formatted as: .

The fully-qualified globally unique id of the skill, formatted as: ..

Parameterizable Behavior Tree Description, if set, the executive can retrieve a tree for this skill's ID.

Information about options supported by this skill during execution.

Display name of the skill.

Whether the skill is sideloaded into the solution.

A complete descriptor set for the skill's parameter and execute_result proto messages.

The filename of the skill image.

A binary proto file that contains the parameterizable behavior tree.

Contains user-set options about what skill features are supported by this skill.

Specifies the runtime dependencies of this skill.

A description of the input parameter message for the skill.

A description of the execute result for a skill invocation.

List of status codes which the specific skill may return. Codes are specific to the skill's ID (.), i.e., different skills can have the same numeric ID representing different errors.

The globally-unique name of this skill instance.

Dispatcher handle to project this skill.

Dispatcher handle to execute this skill.

Maps resource slot name to a handle satisfying the slot's selector. This map must contain an entry for each slot in the implementation's resource_selectors map.

DEPRECATED: Please retrieve default parameters via the ParameterDescription returned by SkillRegistry::GetSkills, and not through a SkillInstance. params and default_parameters are passed here to maintain our current default override handling. Please do not add additional dependencies on theses fields.

The fully-qualified globally unique id of the skill, formatted as: ..

The skill's id.

The skill's vendor.

Documentation about the skill which will be shown to users. This field should contain the description of the skill as a summary.

Contains user-set options about what skill features are supported by this skill.

Specifies the runtime dependencies of this skill.

A description of the input parameter message for Execution of the skill.

A description of the return type for Execution of the skill.

The display name of the skill to be shown in the UI.

Extended documentation about the skill which will be shown to users. This field contains additional information about the skill that is used by automatic skill documentation generation tools. It is optional and can be left empty. However, only skills that specify this field will be included in automatic skill documentation generation.

List of status codes which the specific skill may return. Codes are specific to the skill's ID (.), i.e., different skills can have the same numeric ID representing different errors.

This skill's assets.

The skill's id.

The skill's vendor.

Documentation about the skill which will be shown to users. This field should contain the description of the skill as a summary.

The display name of the skill to be shown in the UI.

Unit of the parameter. This should be the symbol for the unit rather than the full name (e.g., "m" for meters or "rad" for radian). If empty, the parameter is treated as unitless.

When possible, field names should contain their unit as a suffix. This ensures that people working with generated code for the parameter proto always have access to unit information.

Designator for a field that holds an Icon2PositionPart reference. This can be used to designate a particular field in the UI that should be used to select a part name.

Mapping from skill name to registration overrides.

The expected duration of the update.

World updates to apply.

Time at which the update is expected to start.

If omitted, the update is assumed to start immediately at the end of the previous update, or at the beginning of skill execution if there is no previous update.

The id of the behavior tree's skill representation to remove.

The type of volume lock.

The holder has an exclusive lock on this object and is free to change its state. For example, a writer lock must be held to control a robot.

The holder has a shared lock on this object and its state must not change until all reader locks are released.

The holder is allowed to change the contents of the volume. This is an exclusive lock. For example, a robot moving through a swept volume would need to acquire this type of lock.

Unspecified sharing type is treated the same as WRITE_NONEXCLUSIVE.

The holder requires exclusive access to read and change the state of this resource.

The holder can read but may not change the state of this resource.

The holder can read and change the state of this resource. It requires exclusive access to change the state, but does not require exclusive access to read it.

The holder can read and change the state of this resource, but does not require exclusive access to either.