A schema is the data contract for information moving through an agent pipeline. It specifies the fields, structure, and expected form of messages, tool inputs and outputs, observations, and durable artifacts. Shared schemas let modules exchange information predictably and make the path of data traceable across the system. section 2.5.3 section 2.8
Schemas define operational boundaries, not merely storage layouts. A normalization stage converts raw reconnaissance results into a predictable artifact without scoring or recommending action. At that boundary, orchestration can validate the artifact and enforce scope constraints before later agents consume it. section 3.10.2
Treat artifacts as contracts between agents. Silent artifact drift occurs when a field is added, renamed, or subtly changes meaning while downstream stages continue to run. The resulting reports can remain plausible even as priorities degrade or findings disappear. Version schemas, validate explicitly, store schema metadata with data, and fail loudly when an artifact cannot be parsed confidently. section 3.11.1
Versioned reconnaissance artifact
A reconnaissance record can use schema "recon-v1" and pair a host with DNS addresses, response headers, an optional WAF hint, optional TLS data, contextual notes, and a UTC timestamp. This gives later reasoning or reporting stages recorded context rather than requiring them to infer what a probe did or found. section 4.3.2 section 4.4.1
Listing 4.1 Setting the artifact schema for storing structure recon data
from dataclasses import dataclass, asdict, field
from datetime import datetime, timezone
from typing import Dict, List, Optional
import json
import os
# Output path can be overridden via env var
OUT = os.environ.get("RECON_OUT", "recon.jsonl")
@dataclass
class Artifact:
schema: str = "recon-v1"
host: str = ""
a: List[str] = field(default_factory=list) #A
cname: Optional[str] = None #B
headers: Dict[str, str] = field(default_factory=dict) #C
waf_hint: Optional[bool] = None #D
tls: Optional[Dict] = None #E
notes: List[str] = field(default_factory=list) #F
ts: str = "" #G
Inspect a concrete dataclass schema for a versioned reconnaissance artifact, including required defaults, optional evidence fields, and a timestamp. section 4.3.2
The same discipline applies to tools. A tool schema can expose metadata such as a name and description alongside an invoke method that accepts structured input and returns structured output. Aligning those formats with observation schemas supports inspection of data as it crosses agent modules. section 2.5.3





















































