Manifest Schema
The AIM manifest schema defines the structure for agent instruction manifests.
Top-Level Structure
aim: "1.0" # Protocol version (required)
metadata: # Manifest metadata (required)
name: string # Kebab-case identifier
version: string # Semver version
description: string # Brief description
context: # Operational context
persona: string # Agent role/persona
audience: string # Target users
domain: string # Subject domain
environment: string # dev | staging | production
capabilities: [] # What the agent can do
knowledge: [] # Domain knowledge and guidelines
governance: # Rules and constraints
rules: []
guardrails: {}
quality_gates: {}
audit: {}
dependencies: [] # Other manifests this depends on
inherits: [] # Base manifests to extendCapabilities
capabilities:
- name: string # Unique identifier
tags: [string] # Categorization tags
index: string # One-line description (Tier 0)
schema: # Tier 1: Structured interface
inputs: [...]
outputs: [...]
preconditions: [...]
instructions: string # Tier 2: Full how-to
examples: [...] # Usage examples
dispatch: # Tier 3: Execution target
type: mcp | cli | rest | code
config: {...}Governance Rules
governance:
rules:
- name: string
description: string
category: security | quality | architecture | compliance
enforcement: static | semantic | injected
detect:
type: tool | pattern | semantic
# For tool:
command: string
# For pattern:
match: string (regex)
# For semantic:
criteria: string
model: string
threshold: number
action: block | warn | require_approval | log
severity: critical | error | warning | info
message: string
fix_hint: stringFull Schema
See the JSON Schema for the complete specification.