Skip to content
GitHub

AI Skills

lexigram-ai-skills provides a registry-based skill system where agents can discover, validate, and execute composable capabilities. Skills are async functions or classes registered by name, with JSON Schema parameter validation, caching, permission checking, and execution lifecycle management.

For full configuration details, see the lexigram-ai-skills package docs.

1. The Contracts

Section titled “1. The Contracts”

The skill system is built on three core protocols from lexigram.contracts.ai.skills:

from typing import Any, Protocol, runtime_checkable
from lexigram.result import Result
from lexigram.contracts.ai.skills import (
    SkillDefinition, SkillResult, SkillError,
)




class SkillProtocol(Protocol):
    @property
    def definition(self) -> SkillDefinition: ...
    async def execute(self, **kwargs: Any) -> Result[SkillResult, SkillError]: ...
    def validate(self, params: dict[str, Any]) -> list[str]: ...




class SkillRegistryProtocol(Protocol):
    def register(self, skill: SkillProtocol) -> None: ...
    def get(self, name: str) -> SkillProtocol | None: ...
    def list_skills(
        self, category: str | None = None, permissions: list[str] | None = None,
    ) -> list[SkillDefinition]: ...
    def get_schemas(self) -> list[dict[str, Any]]: ...




class SkillExecutorProtocol(Protocol):
    async def execute(
        self,
        skill_name: str,
        params: dict[str, Any],
        user_id: str | None = None,
        session_id: str | None = None,
    ) -> Result[SkillResult, SkillError]: ...

SkillDefinition and SkillResult are the core data types:

from dataclasses import dataclass, field
from typing import Any




@dataclass(frozen=True)
class SkillDefinition:
    name: str
    description: str
    parameters_schema: dict[str, Any] = field(default_factory=dict)
    category: str = "general"
    requires_confirmation: bool = False
    cacheable: bool = False
    timeout_seconds: float = 30.0
    permissions: list[str] = field(default_factory=list)




@dataclass(frozen=True)
class SkillResult:
    skill_name: str
    success: bool
    output: Any = None
    error: str | None = None
    duration_ms: float = 0.0
    cached: bool = False
    metadata: dict[str, Any] = field(default_factory=dict)

2. Configuration

Section titled “2. Configuration”

Add SkillsModule and configure execution defaults:

from lexigram import Application
from lexigram.ai.skills import SkillsModule, SkillsConfig


app = Application(name="my-app")
app.add_module(SkillsModule.configure(
    SkillsConfig(
        enable_builtin=True,
        builtin_skills=["current_datetime", "web_search", "math_calculate"],
        default_timeout_seconds=30.0,
    ),
))
application.yaml
ai_skills:
  name: ai-skills
  default_timeout_seconds: 30.0
  max_retries: 2
  max_concurrent_executions: 10
  cache_enabled: true
  cache_ttl_seconds: 300
  cache_backend: in_memory
  enforce_permissions: false
  auto_discover: false
  enable_builtin: true
  builtin_skills:
    - current_datetime
    - math_calculate
    - web_search
    - web_fetch

3. Defining Skills

Section titled “3. Defining Skills”

Use the @skill decorator to register a class as a skill:

from lexigram.ai.skills import skill




@skill(
    name="generate_summary",
    description="Generate a summary of the given text",
    parameters={
        "type": "object",
        "properties": {
            "text": {"type": "string", "description": "The text to summarize"},
            "max_length": {"type": "integer", "description": "Max summary length"},
        },
        "required": ["text"],
    },
)
class GenerateSummary:
    async def execute(self, text: str, max_length: int = 100) -> str:
        # Implementation
        return text[:max_length] + "..."

For simple functions, use FunctionSkill:

from lexigram.ai.skills import FunctionSkill




async def web_search(query: str, max_results: int = 5) -> list[dict]:
    # Implementation
    return [{"title": "Result", "url": "https://example.com"}]




search_skill = FunctionSkill(
    name="web_search",
    description="Search the web for information",
    execute_fn=web_search,
    parameters={
        "type": "object",
        "properties": {
            "query": {"type": "string"},
            "max_results": {"type": "integer"},
        },
        "required": ["query"],
    },
)

4. Registering and Executing

Section titled “4. Registering and Executing”

Register skills with the registry and execute them through the executor:

from lexigram import Application
from lexigram.ai.skills import SkillsModule, SkillsConfig
from lexigram.contracts.ai.skills import (
    SkillRegistryProtocol,
    SkillExecutorProtocol,
)




async def skill_workflow() -> None:
    async with Application.boot(
        modules=[SkillsModule.configure(SkillsConfig(enable_builtin=True))]
    ) as app:
        registry = await app.container.resolve(SkillRegistryProtocol)
        executor = await app.container.resolve(SkillExecutorProtocol)


        for skill_def in registry.list_skills():
            print(f"{skill_def.name}: {skill_def.description}")


        result = await executor.execute(
            skill_name="current_datetime",
            params={"format": "%Y-%m-%d %H:%M:%S"},
        )
        if result.is_ok():
            sr = result.unwrap()
            print(f"Output: {sr.output}, Duration: {sr.duration_ms}ms")

5. Built-in Skills

Section titled “5. Built-in Skills”

The framework ships with several built-in skills:

SkillDescription
current_datetimeGet the current date and time
math_calculateEvaluate mathematical expressions
web_searchSearch the web for information
web_fetchFetch content from a URL
text_summarizeSummarize text content
text_translateTranslate text between languages
code_executeExecute code in a sandbox

Built-in skills are registered during boot when enable_builtin: true and their names appear in the builtin_skills list.

6. Skill Composition

Section titled “6. Skill Composition”

Chain skills using SkillPipeline:

from lexigram.ai.skills import SkillPipeline




async def composed_workflow() -> None:
    pipeline = SkillPipeline(steps=[
        ("web_search", {"query": "{query}"}),
        ("text_summarize", {"text": "{web_search.output}"}),
    ])
    result = await pipeline.execute(query="latest AI research")

ParallelSkills runs skills concurrently; SkillRouter dispatches based on input.

7. Testing

Section titled “7. Testing”

Use SkillsModule.stub() for isolated tests:

from lexigram import Application
from lexigram.ai.skills import SkillsModule
from lexigram.contracts.ai.skills import SkillExecutorProtocol




async def test_skill_executor_resolves() -> None:
    async with Application.boot(modules=[SkillsModule.stub()]) as app:
        executor = await app.container.resolve(SkillExecutorProtocol)
        assert executor is not None

Next Steps

Section titled “Next Steps”