Agent

An agent is a directory containing an AGENT.md file with YAML frontmatter and an optional Markdown body providing system instructions. Agents are discovered from two locations: workspace .stencila/agents/ and user config ~/.config/stencila/agents/.

Properties

The Agent type has these properties:

Name	Description	Type	Inherited from
`id`	The identifier for this item.	`String`	`Entity`
`alternateNames`	Alternate names (aliases) for the item.	`String`*	`Thing`
`description`	A description of the item.	`String`	`Thing`
`identifiers`	Any kind of identifier for any kind of Thing.	(`PropertyValue` \| `String`)*	`Thing`
`images`	Images of the item.	`ImageObject`*	`Thing`
`name`	The name of the agent.	`String`	-
`url`	The URL of the item.	`String`	`Thing`
`workType`	The type of `CreativeWork` (e.g. article, book, software application).	`CreativeWorkType`	`CreativeWork`
`doi`	The work's Digital Object Identifier (https://doi.org/).	`String`	`CreativeWork`
`about`	The subject matter of the content.	`ThingVariant`*	`CreativeWork`
`abstract`	A short description that summarizes a `CreativeWork`.	`Block`*	`CreativeWork`
`authors`	The authors of the `CreativeWork`.	`Author`*	`CreativeWork`
`provenance`	A summary of the provenance of the content within the work.	`ProvenanceCount`*	`CreativeWork`
`contributors`	A secondary contributor to the `CreativeWork`.	`Author`*	`CreativeWork`
`editors`	People who edited the `CreativeWork`.	`Person`*	`CreativeWork`
`maintainers`	The maintainers of the `CreativeWork`.	(`Person` \| `Organization`)*	`CreativeWork`
`comments`	Comments about this creative work.	`Comment`*	`CreativeWork`
`dateCreated`	Date/time of creation.	`Date`	`CreativeWork`
`dateReceived`	Date/time that work was received.	`Date`	`CreativeWork`
`dateAccepted`	Date/time of acceptance.	`Date`	`CreativeWork`
`dateModified`	Date/time of most recent modification.	`Date`	`CreativeWork`
`datePublished`	Date of first publication.	`Date`	`CreativeWork`
`funders`	People or organizations that funded the `CreativeWork`.	(`Person` \| `Organization`)*	`CreativeWork`
`fundedBy`	Grants that funded the `CreativeWork`; reverse of `fundedItems`.	(`Grant` \| `MonetaryGrant`)*	`CreativeWork`
`genre`	Genre of the creative work, broadcast channel or group.	`String`*	`CreativeWork`
`keywords`	Keywords or tags used to describe this content. Multiple entries in a keywords list are typically delimited by commas.	`String`*	`CreativeWork`
`isPartOf`	An item or other CreativeWork that this CreativeWork is a part of.	`CreativeWorkVariant`	`CreativeWork`
`licenses`	License documents that applies to this content, typically indicated by URL, but may be a `CreativeWork` itself.	(`CreativeWorkVariant` \| `String`)*	`CreativeWork`
`parts`	Elements of the collection which can be a variety of different elements, such as Articles, Datatables, Tables and more.	`CreativeWorkVariant`*	`CreativeWork`
`publisher`	A publisher of the CreativeWork.	`Person` \| `Organization`	`CreativeWork`
`bibliography`	A bibliography of references which may be cited in the work.	`Bibliography`	`CreativeWork`
`references`	References to other creative works, such as another publication, web page, scholarly article, etc.	`Reference`*	`CreativeWork`
`text`	The textual content of this creative work.	`Text`	`CreativeWork`
`title`	The title of the creative work.	`Inline`*	`CreativeWork`
`repository`	URL of the repository where the un-compiled, human readable source of the work is located.	`String`	`CreativeWork`
`path`	The file system path of the source of the work.	`String`	`CreativeWork`
`commit`	The commit hash (or similar) of the source of the work.	`String`	`CreativeWork`
`version`	The version of the creative work.	`String` \| `Number`	`CreativeWork`
`whenToUse`	Positive selection signals describing when this agent should be used.	`String`*	-
`whenNotToUse`	Negative selection signals describing when this agent should not be used.	`String`*	-
`frontmatter`	Frontmatter containing agent metadata.	`String`	-
`content`	The content of the agent (the Markdown body providing system instructions).	`Block`*	-
`models`	Model identifiers for the agent.	`String`*	-
`providers`	Provider identifiers for the agent.	`String`*	-
`modelSize`	Model size preference for the agent.	`String`	-
`reasoningEffort`	Reasoning effort level for the agent.	`String`	-
`historyThinkingReplay`	Whether to replay assistant thinking and reasoning in conversation history.	`String`	-
`truncationPreset`	Named preset for tool output truncation limits.	`String`	-
`compactionTriggerPercent`	Context usage percentage that triggers proactive history compaction.	`UnsignedInteger`	-
`trustLevel`	Trust level controlling how strictly the agent's operations are guarded.	`String`	-
`allowedSkills`	Skill names this agent can use.	`String`*	-
`allowedTools`	Tool names available to the agent.	`String`*	-
`allowedDomains`	Domain allowlist for web_fetch.	`String`*	-
`disallowedDomains`	Domain denylist for web_fetch.	`String`*	-
`enableMcp`	Whether to enable MCP tools.	`Boolean`	-
`enableMcpCodemode`	Whether to enable MCP codemode orchestration.	`Boolean`	-
`allowedMcpServers`	MCP server IDs this agent is allowed to use.	`String`*	-
`maxTurns`	Maximum conversation turns (0 = unlimited).	`Integer`	-
`toolTimeout`	Default timeout for tool execution in seconds.	`Integer`	-
`maxToolRounds`	Maximum tool-call rounds per user input.	`Integer`	-
`maxSubagentDepth`	Maximum subagent nesting depth.	`Integer`	-
`compatibility`	Environment requirements for the agent.	`String`	-