Core Tasks

Atomic task definitions and execution protocols

Core Tasks

Core tasks are atomic, reusable building blocks for workflows. Each task follows the HO-TP anatomy pattern.

Task Anatomy

Every task includes:

task_name · string · required

Human-readable task name

task_id · string · required

Unique task identifier (e.g., TSK-020)

execution_type · string · required

Execution mode: Agent, Script, Manual, Hook

responsible_executor · string · required

Who/what executes: @jarvis, @agent:{id}, @human, @hook:{name}

input · object · required

Required input parameters and types

output · object · required

Expected output format and location

acceptance_criteria · array · required

Checklist for task completion validation

---

Available Tasks

extract-dna

Extract cognitive DNA from insights and themes.

Task ID: TSK-020

Execution:

• Type: Agent
• Executor: @jarvis

Inputs:

insights · json · required

Extracted insights from content

themes · array · required

Theme classifications

source_id · string · required

Source identifier (e.g., “CG”, “JM”)

chunks · array · required

Content chunks for reference

Outputs:

dna_config · yaml

5-layer DNA structure saved to agents/persons/{name}/DNA-CONFIG.yaml

layer_counts · json

Count of items per DNA layer (L1-L5)

Phases:

Show Phase 1: Philosophy Extraction (L1)

Quality Gate: QG-DNA-001

1. Identify core beliefs and worldview statements
2. Extract value system indicators
3. Map to PHILOSOPHIES layer

Show Phase 2: Mental Models (L2)

Quality Gate: QG-DNA-002

1. Identify thinking frameworks
2. Extract decision-making patterns
3. Map to MENTAL-MODELS layer

Show Phase 3: Heuristics (L3)

Quality Gate: QG-DNA-003

1. Extract practical rules
2. Identify decision shortcuts
3. Map to HEURISTICS layer

Show Phase 4: Frameworks (L4)

Quality Gate: QG-DNA-004

1. Identify structured methodologies
2. Extract process definitions
3. Map to FRAMEWORKS layer

Show Phase 5: Methodologies (L5)

Quality Gate: QG-DNA-005

1. Extract step-by-step implementations
2. Identify tactical procedures
3. Map to METHODOLOGIES layer

Acceptance Criteria:

• All 5 DNA layers populated
• Minimum 3 items per layer
• Source citations for each item (^[chunk_id])
• Valid YAML structure

Handoff:

• Next Task: trigger-agent when dna_complete
• Data Passed: DNA-CONFIG path, layer_counts

Source: core/tasks/extract-dna.md:1-115

---

detect-role

Detect job roles and functions mentioned in content.

Task ID: TSK-012

Execution:

• Type: Script
• Executor: @script:role_detector.py

Inputs:

text · string · required

Content to analyze for role mentions

source_id · string

Source identifier for tracking

Detection Levels:

1. Direct (weight: 1.0) - Role name appears explicitly
2. Inferred (weight: 0.7) - Activity implies role (e.g., “writing sales copy” → COPYWRITER)
3. Emergent (weight: 0.5) - Unknown role detected via generic patterns

Outputs:

roles_detected · array

List of detected roles with detection type, weight, and context

new_roles · array

Emergent roles not in vocabulary (candidates for agent creation)

triggers_activated · array

Roles that crossed threshold for agent creation

Registry Update:

• Increments mention_count in ENTITY-REGISTRY.json
• Updates mention_breakdown (direct/inferred/emergent)
• Calculates weighted_score
• Logs detection in detection_history

Trigger Thresholds:

• Established: weighted_score >= 10 AND sources >= 2
• Emerging: weighted_score >= 5 AND sources >= 1

Source: core/tasks/detect-role.md:1-17, core/intelligence/role_detector.py:1-856

---

normalize-entities

Canonicalize entity names (persons, themes, roles, concepts).

Task ID: TSK-008

Execution:

• Type: Script
• Executor: @script:entity_normalizer.py

Inputs:

name · string · required

Raw entity name to normalize

entity_type · string · required

Type: person, theme, role, concept

source_id · string

Source that mentioned this entity

domain_hint · string

Domain context for boost matching (e.g., “SALES”)

Resolution Order:

1. Exact match on canonical names (score: 1.0)
2. Alias match on known aliases (score: 1.0)
3. Fuzzy match via difflib (threshold: 0.85)
4. Domain boost for same-domain entities (+0.10 bonus)
5. Taxonomy check against DOMAINS-TAXONOMY.yaml
6. New entity creation if no match found

Auto-Merge Rules:

• Score ≥ 0.95: Auto-merge, add as alias
• Score 0.85-0.94: Add to review queue
• Score < 0.85: Create new entity

Outputs:

canonical · string

Canonical entity name

match_type · string

Match method: exact, alias, fuzzy, domain_boost, taxonomy, new

score · number

Similarity score (0.0-1.0)

created · boolean

Whether new entity was created

Canonical Name Formats:

• Person: Title Case (“Alex Hormozi”)
• Role: UPPER-CASE-WITH-HYPHENS (“SALES-MANAGER”)
• Theme: lowercase-with-hyphens (“processo-vendas”)
• Concept: Title Case (“Farm System”)

Source: core/tasks/normalize-entities.md:1-16, core/intelligence/entity_normalizer.py:1-566

---

analyze-themes

Detect and classify themes in content.

Task ID: TSK-015

Execution:

• Type: Script
• Executor: @script:theme_analyzer.py

Inputs:

text · string · required

Content to analyze for themes

source_id · string

Source identifier

Theme Categories:

Code	Theme	Description
01	ESTRUTURA-TIME	Team structure, roles
02	PROCESSO-VENDAS	Sales process, techniques
03	CONTRATACAO	Hiring, onboarding
04	COMISSIONAMENTO	Compensation structures
05	METRICAS	Performance metrics
06	FUNIL-APLICACAO	Lead qualification
07	PRICING	Pricing strategy
08	FERRAMENTAS	Tools, tech stack
09	GESTAO	Management practices
10	CULTURA-GAMIFICACAO	Culture, motivation
99	SECUNDARIO	Other insights

Detection Logic:

Uses keyword patterns and context analysis:

if mentions("roles", "team", "structure"):
    theme = "01-ESTRUTURA-TIME"
elif mentions("sales", "closing", "objection"):
    theme = "02-PROCESSO-VENDAS"
elif mentions("hire", "onboard", "recruit"):
    theme = "03-CONTRATACAO"
# ... etc

Outputs:

themes_detected · array

List of theme IDs with confidence scores

cross_references · array

Themes with secondary relevance

Source: core/tasks/analyze-themes.md:1-17

---

process-batch

Process multiple files in batch with checkpointing.

Task ID: TSK-025

Execution:

• Type: Agent
• Executor: @jarvis

Inputs:

file_list · array · required

List of file paths to process

workflow_id · string · required

Workflow to execute (e.g., “wf-pipeline-full”)

checkpoint_interval · number

Files per checkpoint (default: 1)

Behavior:

• Processes files sequentially
• Creates checkpoint after each file
• Auto-resume on interruption
• Aggregates metrics across batch

Outputs:

batch_summary · object

Total files, success count, errors, processing time

checkpoints · array

List of checkpoint IDs created

Source: core/tasks/process-batch.md:1-25

---

validate-cascade

Validate post-processing cascade completeness.

Task ID: TSK-030

Execution:

• Type: Script
• Executor: @script:validate_cascade.py

Inputs:

source_id · string · required

Source identifier to validate

Validation Checks:

1. Agent Memories Updated - All relevant agents have source in memory
2. Dossiers Created - Persons/themes have dossiers
3. Knowledge Base Populated - Insights written to correct categories
4. Registry Consistency - ENTITY-REGISTRY matches artifacts
5. Cross-References Valid - All links resolve correctly

Outputs:

validation_result · object

Pass/fail status per check with details

issues_found · array

List of issues with severity (CRITICAL, WARNING, INFO)

Source: core/tasks/validate-cascade.md:1-21

---

Task Registry

All available tasks are indexed in core/tasks/TASK-REGISTRY.md:

| Task ID | Name | Executor | Status |
|---------|------|----------|--------|
| TSK-020 | extract-dna | @jarvis | active |
| TSK-012 | detect-role | @script | active |
| TSK-008 | normalize-entities | @script | active |
| TSK-015 | analyze-themes | @script | active |
| TSK-025 | process-batch | @jarvis | active |
| TSK-030 | validate-cascade | @script | active |

---

Creating New Tasks

Task Template

---
task: TSK-XXX
execution_type: Agent|Script|Manual|Hook
responsible: @jarvis|@agent:{id}|@human|@hook:{name}
---

## Task Anatomy

| Field | Value |
|-------|-------|
| task_name | Your Task Name |
| status | active|deprecated |
| responsible_executor | See above |
| execution_type | See above |
| input | {json object} |
| output | {json object} |
| action_items | {list} |
| acceptance_criteria | {checklist} |

## Inputs

[Document inputs]

## Outputs

[Document outputs]

## Execution

[Step-by-step instructions]

## Acceptance Criteria

- [ ] Criterion 1
- [ ] Criterion 2

## Handoff

| Next Task | Trigger | Data Passed |
|-----------|---------|-------------|
| task-id | condition | data |

Conventions

1. IDs: Use TSK-### format (sequential)
2. Files: Use kebab-case.md in core/tasks/
3. Atomicity: One clear responsibility per task
4. Idempotency: Safe to re-run without side effects
5. Testability: Clear acceptance criteria

---

See Also

• Workflows - Task orchestration
• Intelligence Layer - Python scripts
• Schemas - Data validation