Agent Prompts

System prompts, task prompts, and instruction sets -- the intelligence layer powering every AI interaction

16 prompts
prompt-system-clinical-voice.md
type: system_prompt
model: claude-sonnet-4-6
tokens: 580
temperature: 0.3

Clinical Voice System Prompt

Defines the clinical voice persona for all patient-facing AI interactions. Ensures tone consistency across every therapeutic touchpoint.

Voice Attributes:
- Warm but professional
- Evidence-informed language
- Trauma-aware framing
- Person-first terminology
- Strengths-based perspective
system-promptvoiceclinical
prompt-system-content-writer.md
type: system_prompt
tokens: 720
temperature: 0.6

Content Writer System Prompt

Establishes the content creation persona for blog posts, articles, and long-form written assets. Balances SEO awareness with authentic voice.

Capabilities:
- Long-form article drafting
- SEO-optimized headings
- Internal link suggestions
- Voice-matched tone control
system-promptcontentblog
prompt-task-session-note.md
type: task_prompt
input: session_transcript
output: SOAPNote
tokens: 420
temperature: 0.2

Session Note Generator

Transforms raw session transcripts into structured SOAP notes with clinical precision. Extracts subjective reports, objective observations, assessments, and treatment plans.

interface SOAPNote {
  subjective: string;
  objective: string;
  assessment: string;
  plan: string[];
  risk_level: "low" | "moderate" | "high";
}
task-promptsoapclinical
prompt-task-dbt-skill-match.md
type: task_prompt
input: client_state
output: SkillRecommendation
temperature: 0.3

DBT Skill Matcher

Analyzes client emotional state and context to recommend the most relevant DBT skills from all four modules: Mindfulness, Distress Tolerance, Emotion Regulation, and Interpersonal Effectiveness.

interface SkillRecommendation {
  module: string;
  skill_name: string;
  rationale: string;
  practice_steps: string[];
  contraindications?: string[];
}
task-promptdbtskill-matching
prompt-task-crisis-assess.md
type: task_prompt
priority: critical
output: CrisisAssessment
temperature: 0.1
CRITICAL PRIORITY

Crisis Assessment Prompt

Evaluates crisis indicators with maximum determinism. Outputs structured risk assessment with immediate action recommendations.

interface CrisisAssessment {
  risk_level: "acute" | "high" | "moderate";
  indicators: string[];
  immediate_actions: string[];
  escalation_path: string;
  safety_plan: boolean;
}

Always escalate. Never minimize.

task-promptcrisiscritical
prompt-template-blog-article.md
type: template
output: BlogArticle
tokens: 650
temperature: 0.7

Blog Article Template

Structured template for generating long-form blog content with SEO metadata, structured headings, and embedded CTAs.

interface BlogArticle {
  title: string;
  meta_description: string;
  slug: string;
  sections: Section[];
  cta: CallToAction;
  tags: string[];
}
templateblogcontent
prompt-template-social-linkedin.md
type: template
platform: linkedin
tokens: 380
temperature: 0.6

LinkedIn Post Template

Generates scroll-stopping LinkedIn posts with hook-first structure, strategic line breaks, and engagement-optimized formatting.

Structure:
1. Hook (1 line, pattern interrupt)
2. Context (2-3 lines)
3. Insight / Story (3-5 lines)
4. Takeaway (1-2 lines)
5. CTA + Hashtags
templatelinkedinsocial
prompt-template-newsletter.md
type: template
platform: substack
tokens: 520
temperature: 0.5

The Resilience Letter Template

Newsletter template for Substack editions. Combines clinical insight with personal narrative, structured for reader trust and retention.

Sections:
- Opening Reflection
- Clinical Insight
- Real-World Application
- Guided Exercise
- Weekly Challenge
- Resources + Links
templatenewslettersubstack
prompt-chain-content-pipeline.md
type: chain
steps: 4
model: claude-sonnet-4-6

Content Pipeline Chain

Four-step sequential chain that takes a topic seed and produces publish-ready content through progressive refinement.

1Ideation
->
2Outline
->
3Draft
->
4Polish
chaincontentpipeline
prompt-chain-intake-to-plan.md
type: chain
steps: 3

Intake-to-Plan Chain

Three-step clinical chain: intake assessment, diagnostic formulation, and treatment plan generation with built-in human review gates.

1Intake Assessment
->
2Diagnostic Formulation
->
3Treatment Plan

Mandatory human review gate between steps 2 and 3.

chainclinicalintake
prompt-few-shot-emotion-label.md
type: few_shot
examples: 3
temperature: 0.2

Emotion Labeling Few-Shot

Three-example few-shot prompt for classifying client emotional states from session transcript excerpts. Maps statements to primary and secondary emotion labels.

Example 1:
Input:  "I just feel like nothing matters"
Output: { primary: "hopelessness",
          secondary: "numbness",
          intensity: 8 }

Example 2:
Input:  "I snapped at my partner again"
Output: { primary: "anger",
          secondary: "guilt",
          intensity: 7 }
few-shotemotionlabeling
prompt-few-shot-validation.md
type: few_shot
examples: 6
temperature: 0.3

DBT Validation Levels Few-Shot

Six examples covering all six levels of DBT validation, from active listening through radical genuineness. Teaches the model to match validation level to clinical context.

Levels:
L1: Active Listening
L2: Accurate Reflection
L3: Articulating the Unverbalized
L4: Historical Validation
L5: Normalizing
L6: Radical Genuineness
few-shotdbtvalidation
prompt-task-hypnosis-script.md
type: task_prompt
output: HypnosisScript
tokens: 480
temperature: 0.4

Hypnosis Script Generator

Generates clinically-informed hypnosis scripts tailored to client goals. Includes induction, deepening, suggestion, and emergence phases.

interface HypnosisScript {
  induction: string;
  deepening: string;
  therapeutic_suggestions: string[];
  post_hypnotic: string;
  emergence: string;
  duration_minutes: number;
}
task-prompthypnosisclinical
prompt-task-group-prep.md
type: task_prompt
output: GroupSessionPlan
tokens: 440
temperature: 0.4

Group Session Prep

Prepares structured group therapy session plans for IOP settings. Includes warm-up, psychoeducation, experiential exercise, and closing components.

interface GroupSessionPlan {
  topic: string;
  warm_up: Activity;
  psychoeducation: string;
  experiential: Activity;
  process_questions: string[];
  closing: string;
  materials: string[];
}
task-promptgroup-therapyiop
prompt-template-diary-card.md
type: template
output: DBTDiaryCard
tokens: 350
temperature: 0.2

DBT Diary Card Prompt

Generates structured daily diary card entries for DBT clients. Tracks emotions, urges, skills used, and medication compliance in a standardized format.

interface DBTDiaryCard {
  date: string;
  emotions: EmotionEntry[];
  urges: UrgeEntry[];
  skills_used: string[];
  medication_taken: boolean;
  self_care: string[];
  notes: string;
}
templatedbtdiary-card
prompt-system-brand-guardian.md
type: system_prompt
tokens: 890
temperature: 0.2

Brand Guardian System Prompt

Enforces brand voice consistency and compliance across all generated content. Acts as a final review layer before any content is published.

Enforcement Rules

  • All content must use person-first language
  • Clinical claims require citation or qualifier
  • Brand voice score must exceed 0.85 threshold
  • No absolute guarantees about treatment outcomes
  • Terminology must match approved glossary

Rejection Criteria

  • Stigmatizing language detected
  • Unsupported clinical claims
  • Off-brand tone or voice deviation
  • Missing required disclaimers
system-promptbrandcompliance

Edward Flynn

LPC, LCADC, ACS, CHt

Architect of the intelligence layer powering clinical AI interactions. These prompts encode decades of therapeutic expertise into structured instructions that guide every agent conversation, content generation, and clinical decision-support workflow across the Windfield IO platform.