Agentic AI Atlasby a5c.ai
OverviewWikiGraphFor AgentsEdgesSearchWorkspace
/
GitHubDocsDiscord
iiRecord
Agentic AI Atlas · assimilation/harness/cursor
lib-process:meta--cursora5c.ai
Search record views/
Record · tabs

Available views

II.Record viewspp. 1 - 1
overviewjsongraph
II.
LibraryProcess overview

lib-process:meta--cursor

Reference · live

assimilation/harness/cursor overview

Orchestrate babysitter SDK integration into Cursor IDE/CLI. Cursor is a VS Code fork with built-in AI capabilities. It has: - A CLI (`cursor`) with headless agent mode (`cursor agent --print -p "prompt"`) - Hooks system (1.7+): beforeSubmitPrompt, beforeShellExecution, beforeMCPExecution, afterFileEdit, afterAgentResponse, stop — configured via .cursor/hooks.json - NOTE: As of writing, hooks only fire in IDE mode, NOT in headless CLI mode. The sessionStart hook fires in headless, but afterAgentResponse and stop do not. This is a known limitation that may change — the research phase must verify current state. - Config: .cursor/ directory, .cursorrules for rules - MCP server support for tool access - VS Code extension model (not a standalone plugin system) - Already has minimal KNOWN_HARNESSES entry (HeadlessPrompt only, no callerEnvVars) Key challenge: hooks don't work in headless CLI mode, so the continuation mechanism is limited. The process must research the current state of Cursor hooks in CLI mode and determine whether stop-hook, in-turn, or a hybrid approach is appropriate.

LibraryProcessOutgoing · 7Incoming · 0

Attributes

displayName
assimilation/harness/cursor
description
Orchestrate babysitter SDK integration into Cursor IDE/CLI. Cursor is a VS Code fork with built-in AI capabilities. It has: - A CLI (`cursor`) with headless agent mode (`cursor agent --print -p "prompt"`) - Hooks system (1.7+): beforeSubmitPrompt, beforeShellExecution, beforeMCPExecution, afterFileEdit, afterAgentResponse, stop — configured via .cursor/hooks.json - NOTE: As of writing, hooks only fire in IDE mode, NOT in headless CLI mode. The sessionStart hook fires in headless, but afterAgentResponse and stop do not. This is a known limitation that may change — the research phase must verify current state. - Config: .cursor/ directory, .cursorrules for rules - MCP server support for tool access - VS Code extension model (not a standalone plugin system) - Already has minimal KNOWN_HARNESSES entry (HeadlessPrompt only, no callerEnvVars) Key challenge: hooks don't work in headless CLI mode, so the continuation mechanism is limited. The process must research the current state of Cursor hooks in CLI mode and determine whether stop-hook, in-turn, or a hybrid approach is appropriate.
libraryPath
library/specializations/meta/assimilation/harness/cursor.js
specialization
meta

Outgoing edges

lib_applies_to_domain1
  • domain:software-engineering·DomainSoftware Engineering
lib_belongs_to_specialization1
  • specialization:meta·SpecializationMeta / Process Engineering
lib_covers_topic2
  • topic:developer-experience·TopicDeveloper Experience (DX)
  • topic:integrations·TopicIntegrations
lib_involves_role1
  • role:platform-engineer·Role
lib_requires_skill_area2
  • skill-area:ai-agent-development·SkillAreaAI Agent Development
  • skill-area:orchestration-loop·SkillAreaOrchestration Loop Engineering

Incoming edges

None.

Related pages

No related wiki pages for this record.

Shortcuts

Open in graph
Browse node kind