AI Agents model concepts

What each agent calls the same concept — Kiro · Claude Code · Gemini CLI · Junie · Antigravity

Kiro Claude Code Gemini CLI Junie Antigravity
Concept Kiro Claude Code Gemini CLI Junie Antigravity
Context — persistent project knowledge the agent always carries
Project context
Stack, structure, conventions
.kiro/steering/tech.md
structure.md
product.md
✓ portable Markdown
CLAUDE.md
/memory refs
✓ portable Markdown
GEMINI.md
@file.md imports
✓ portable Markdown
.junie/guidelines.md
✓ portable Markdown
.antigravity/context/
✓ portable Markdown
Global / team context
Org-wide defaults across projects
~/.kiro/steering/
✓ MDM-deployable
~/.claude/CLAUDE.md
✓ global file
~/.gemini/GEMINI.md
✓ global file
IDE settings path
~ per-machine
~/.antigravity/
✓ global dir
Universal bridge
Single file all agents read
AGENTS.md ✓ AGENTS.md ✓ AGENTS.md (via config) AGENTS.md ✓ AGENTS.md ✓
Rules — behavioral constraints and coding standards
Coding rules
Style, patterns, anti-patterns
steering/code-conventions.md
✓ plain MD
CLAUDE.md rules section
.claude/settings.json
✓ plain MD
GEMINI.md rules section
✓ plain MD
.junie/guidelines.md
✓ plain MD
.antigravity/rules/
✓ plain MD
Security policies
Auth rules, input validation
steering/security-policies.md CLAUDE.md section GEMINI.md section guidelines.md section rules/security.md
Permissions / allow-list
What agent can run without asking
hooks config + IDE settings
✗ Kiro-specific
.claude/settings.json allowedTools
✗ Claude-specific
settings.json coreTool
✗ Gemini-specific
Action Allowlist (IDE)
✗ Junie-specific
—
Specs — structured feature planning artifacts
Requirements
User stories, acceptance criteria
.kiro/specs/<feat>/requirements.md
✓ EARS notation, portable MD
CLAUDE.md + /spec command
or spec-kit spec.md
~ via spec-kit
spec-kit spec.md
~ via spec-kit
task prompt / guidelines
~ manual
SKILL.md brainstorming
~ skill-driven
Design doc
Architecture, sequence diagrams
.kiro/specs/<feat>/design.md
✓ portable MD
spec-kit plan.md
or manual docs/
~ via spec-kit
spec-kit plan.md
~ via spec-kit
manual docs/
~ manual
SKILL.md architecture skill
~ skill-driven
Task list / plan
Ordered implementation steps
.kiro/specs/<feat>/tasks.md
✓ portable MD
TodoWrite tool
spec-kit tasks.md
~ session-scoped
Plan Mode proposals
spec-kit tasks.md
~ session-scoped
Junie task panel
✗ IDE-native only
Backlog.md workflows
~ via DevoxxGenie
Skills — reusable playbooks for recurring tasks
Skills / playbooks
Invoke-able instruction sets
.kiro/steering/ (always-on)
powers/ (on-demand)
~ powers = Kiro-native
.claude/skills/SKILL.md
✓ SKILL.md standard
.gemini/skills/SKILL.md
✓ same SKILL.md standard
guidelines.md sections
~ no dedicated format
~/.antigravity/skills/ SKILL.md
✓ same SKILL.md standard
Slash commands
Explicit skill invocation
/spec /steering custom .claude/commands/*.md
/skill-name
@skill-name or /cmd prompt panel only
✗ no custom commands
/skill-name
Automation — agent actions triggered by events
Event-driven hooks
On save / create / delete
.kiro/hooks/*.yaml
✗ Kiro-native format
.claude/hooks/ PreToolUse PostToolUse
✗ Claude-native format
Extensions / MCP triggers
✗ Gemini-native
Agent hooks (planned)
✗ not yet available
workflow chains
~ skill-triggered
Tool / MCP integration
External APIs, databases
.kiro/powers/ + POWER.md .claude/mcp.json settings.json mcpServers .junie/mcp/ MCP via config
Ignore — what the agent should not touch
Ignore files
Exclude paths from agent scope
.gitignore respected .claudeignore .geminiignore .aiignore .gitignore / .aiignore