R.I.S.C.E.A.R. Deep Dive for Researchers¶

This document provides a formal description of the R.I.S.C.E.A.R. persona specification framework, its dimension profiling methodology, and guidance for using the FCC event replay system for reproducible multi-agent research.

1. The 10-Component Specification¶

R.I.S.C.E.A.R. is a structured specification language for defining AI agent personas. Each persona is fully determined by ten components:

#	Component	Field	Type	Description
1	Role	`role`	`str`	The identity and function assigned to the agent. Describes what the persona does within the FCC workflow.
2	Inputs	`inputs`	`list[str]`	Required data, facts, and background information the persona needs to perform its role.
3	Style	`style`	`str`	Communication conventions: tone, language register, formatting rules, and presentation guidelines.
4	Constraints	`constraints`	`list[str]`	Boundaries, limitations, and mandatory rules that govern the persona's output.
5	Expected Output	`expected_output`	`list[str]`	The structure, format, and detail level of artifacts the persona must produce.
6	Archetype	`archetype`	`str`	The fundamental behavioral model the persona embodies (e.g., "The Methodical Organizer").
7	Responsibilities	`responsibilities`	`list[str]`	Ongoing duties and ethical commitments beyond immediate deliverables.
8	Role Skills	`role_skills`	`list[str]`	Specific competencies required for effective execution.
9	Role Collaborators	`role_collaborators`	`list[str]`	Upstream and downstream interaction partners, referenced by persona ID.
10	Role Adoption Checklist	`role_adoption_checklist`	`list[str]`	Validation criteria that must be satisfied before the persona is considered operational.

Formal Representation¶

A persona P is defined as a tuple:

P = (R, I, S, C, E, A, Resp, Skills, Collab, Checklist)

where each component maps to a field in the RISCEARSpec dataclass (src/fcc/personas/models.py). The specification is loaded from YAML and instantiated as a frozen (immutable) dataclass, ensuring referential integrity throughout the system.

Implementation Reference¶

from fcc.personas.models import RISCEARSpec, PersonaSpec

# Load a persona from the registry
from fcc.personas.registry import PersonaRegistry

registry = PersonaRegistry.from_data_dir("src/fcc/data/personas")
persona = registry.get("RC")  # Research Crafter

# Access R.I.S.C.E.A.R. components
spec = persona.riscear
print(f"Role: {spec.role}")
print(f"Archetype: {spec.archetype}")
print(f"Inputs: {spec.inputs}")
print(f"Constraints: {spec.constraints}")
print(f"Expected Output: {spec.expected_output}")
print(f"Responsibilities: {spec.responsibilities}")
print(f"Skills: {spec.role_skills}")
print(f"Collaborators: {spec.role_collaborators}")
print(f"Adoption Checklist: {spec.role_adoption_checklist}")

2. Behavioral Profiling: Discernment Matrix and Design Target Factors¶

Beyond the functional R.I.S.C.E.A.R. specification, each persona is characterized by two behavioral models.

2.1 Discernment Matrix¶

Six traits rated across seven dimensions:

Trait	Construct
Humility	Acknowledgment of biases, limitations, and others' perspectives
Professional Background	Domain expertise and professional context
Curiosity	Drive to explore and consider new perspectives
Taste	Refined judgment and aesthetic sensibility
Inclusivity	Respect for diverse beliefs, cultures, and experiences
Responsibility	Application of discernment for equitable outcomes

2.2 Design Target Factors¶

Six interpersonal factors modeled on the "Super Connector" archetype:

Factor	Construct
Optimism	Technology and connectivity as tools for positive change
Social Connectivity	Leveraging relationships via networks
Influence	Acting as a catalyst for action within professional networks
Diversity Appreciation	Valuing diverse cultures, thoughts, and people
Curiosity	Lifelong learning and intellectual openness
Leadership	Entrepreneurial mindset and natural leadership

2.3 Seven Rating Dimensions¶

Both the Discernment Matrix and Design Target Factors use a shared seven-dimension rating model:

Dimension	Description
`self_rating`	Self-assessment by the persona
`peer_rating`	Assessment by collaborating personas
`survey_rating`	Aggregated survey-based evaluation
`individual_weighted_rating`	Weighted composite of individual assessments
`org_rating`	Organizational-level evaluation
`external_rating`	Assessment from external stakeholders
`ranked_percentile_rating`	Normalized percentile ranking across the ecosystem

This multi-rater model is implemented as the RatingDimensions frozen dataclass.

3. Persona Dimension Profiling Methodology¶

The deepest level of persona specification is the 56-dimension profile organized into 9 categories:

#	Category	Dimensions	Focus
1	Core Persona Elements	7	Agent profile, organizational role, decision authority
2	Behavioral and Motivational Factors	6	Tool adoption, framework preferences, risk tolerance
3	Communication and Learning Styles	4	Channels, information sources, learning preferences
4	Cultural and Social Influences	4	Operational heritage, protocol proficiency, platform engagement
5	Decision-Making and Leadership Approaches	5	Decision style, problem-solving, conflict resolution
6	Professional Development and Wellness	5	Mentorship, growth, sustainability, cross-project mobility
7	Market and Regulatory Awareness	5	Trends, competition, regulations, ethics
8	Innovative Persona Elements	10	Output trace analysis, innovation rate, crisis management
9	Advanced Persona Attributes	10	Ecosystem role, resource budget, RACI, data governance

Interpretation Guide¶

14 of the 56 dimensions were originally designed for consumer persona modeling and have been reinterpreted for AI documentation agents. For example:

Original Dimension	AI Agent Interpretation
Demographic Information	Agent Profile
Purchasing Behavior	Tool/Resource Adoption Patterns
Income Level	Resource Budget / Compute Allocation

The full mapping with rationale is in data/personas/dimension_interpretation_guide.yaml.

Accessing Dimension Profiles¶

from fcc.personas.registry import PersonaRegistry

registry = PersonaRegistry.from_data_dir("src/fcc/data/personas")
persona = registry.get("RC")

if persona.dimension_profile:
    profile = persona.dimension_profile
    for cat_name in profile.CATEGORY_NAMES:
        dims = getattr(profile, cat_name)
        print(f"{cat_name}: {len(dims)} dimensions")
        for dim in dims[:2]:  # Show first 2
            print(f"  - {dim.name}: {dim.description[:60]}...")

4. Reproducible Research with Event Replay¶

The FCC messaging system provides a complete audit trail suitable for reproducible multi-agent research.

4.1 Recording Events¶

from fcc.messaging.bus import EventBus
from fcc.messaging.serialization import EventSerializer

bus = EventBus()
bus.start_recording()

# Run your experiment (simulation, action execution, collaboration session)
# ... all events are automatically captured ...

bus.stop_recording()
history = bus.get_history()

# Persist the event log
EventSerializer.save(history, "experiment_events.json")
print(f"Recorded {len(history)} events")

4.2 Replaying for Verification¶

Another researcher can load the event log and replay it to verify findings:

from fcc.messaging.bus import EventBus
from fcc.messaging.serialization import EventSerializer, EventReplay

# Load the recorded events
events = EventSerializer.load("experiment_events.json")

# Set up analysis subscribers
persona_activations = []
gate_results = []

bus = EventBus()
bus.subscribe(lambda e: persona_activations.append(e)
    if e.event_type.value.startswith("persona.") else None)
bus.subscribe(lambda e: gate_results.append(e)
    if e.event_type.value.startswith("governance.") else None)

# Replay
replayer = EventReplay(bus)
total = replayer.replay(events)
print(f"Replayed {len(events)} events, {total} subscriber deliveries")

4.3 Filtered Replay for Hypothesis Testing¶

Replay only events matching specific criteria to isolate experimental conditions:

# Replay only events from a specific simulation run
replayer.replay_filtered(
    events,
    correlation_id="experiment-run-001",
)

# Replay only events from the ActionEngine
replayer.replay_filtered(
    events,
    source="ActionEngine",
)

4.4 Session Replay¶

Collaboration sessions can be replayed as structured event sequences:

from fcc.collaboration.recording import SessionRecorder
from fcc.collaboration.models import CollaborationSession

session = SessionRecorder.load_json("session_data.json")

bus = EventBus()
analysis_events = []
bus.subscribe(lambda e: analysis_events.append(e))

deliveries = SessionRecorder.replay_session(session, bus)
print(f"Session {session.session_id}: {len(session.turns)} turns, "
      f"{deliveries} event deliveries")

5. Cross-Reference Matrix for Interaction Analysis¶

The CrossReferenceMatrix enables systematic analysis of persona-to-persona interaction patterns:

from fcc.personas.cross_reference import CrossReferenceMatrix

# Load from YAML
# matrix = CrossReferenceMatrix.from_yaml("data/personas/cross_reference.yaml")

# Or auto-generate from persona collaboration links
# matrix = CrossReferenceMatrix.from_personas(registry)

# Query interaction patterns
# upstream = matrix.upstream("BC")   # Who feeds into Blueprint Crafter?
# downstream = matrix.downstream("RC")  # Where does Research Crafter's output go?
# peers = matrix.peers("DE")  # Who collaborates laterally with Doc Evangelist?

6. Citation Information¶

When referencing the FCC framework or R.I.S.C.E.A.R. specification in academic work:

Suggested Citation

INFORMATION COLLECTIVE, LLC. (2026). FCC Agent Team Extension: A Framework for Multi-Agent Documentation Workflows with R.I.S.C.E.A.R. Persona Specifications (Version 0.5.0) [Software]. GitHub. https://github.com/rollingthunderfourtytwo-afk/l2_fcc_agent_team_ext

BibTeX¶

@software{fcc_agent_team_2026,
  author       = {{INFORMATION COLLECTIVE, LLC}},
  title        = {{FCC Agent Team Extension: A Framework for Multi-Agent
                   Documentation Workflows with R.I.S.C.E.A.R. Persona
                   Specifications}},
  year         = {2026},
  version      = {0.5.0},
  url          = {https://github.com/rollingthunderfourtytwo-afk/l2_fcc_agent_team_ext},
  license      = {MIT}
}

Key Framework Attributes for Reporting¶

When describing the framework in a methods section, include:

Persona count: 102 core + 45 vertical + 23 plugin personas across 20 core categories and 6 vertical packs (170 total; 147 core+vertical)
Specification model: 10-component R.I.S.C.E.A.R.
Behavioral profiling: 6-trait Discernment Matrix + 6-factor Design Target Factors, each rated on 7 dimensions
Dimension profiling: 9 categories, 56 dimensions
Workflow graphs: 5-node (base), 20-node (extended), 24-node (complete), 55-node (extended_84)
Action types: 6 (scaffold, refactor, debug, test, compare, document)
Event types: 25 across 8 categories
Quality gates: 25 across all persona categories

Next Steps¶

Reproducibility Guide -- Extended reproducibility practices
Research Methodology -- Using FCC in research settings
Understanding the Event Bus -- Event replay mechanics