cbt-programs

name	cbt-programs
description	Discover, recommend, translate, and co-create open-source CBT programs for different health problems. Use when users ask for CBT program suggestions, multilingual program content, program provenance labels (AI-created/human-created/human-reviewed/healthcare-expert-verified), or want to create and publish a new CBT program in the repository.

CBT Programs Skill

Use this skill for open-source CBT program operations backed by the cbt-programs CLI and data/programs/ registry.

Core capabilities

• Discover CBT programs by condition and language
• Recommend programs based on user-stated problems
• Show trust/provenance labels:
  • AI-CREATED
  • HUMAN-CREATED
  • HUMAN-REVIEWED
  • EXPERT-VERIFIED
• Scaffold new programs for user-specific needs
• Improve and edit existing programs with tracked updates
• Track updates in markdown progress logs
• Generate agent prompts for guided support sessions

Fast workflow

1. Discover candidates:

cbt-programs list --lang en
cbt-programs search "sleep anxiety"
cbt-programs recommend --problem "difficulty sleeping" --lang en --limit 3

2. Inspect one program:

cbt-programs show cbt-insomnia --lang en

3. Validate trust/readiness:

cbt-programs validate cbt-insomnia

4. Create user-specific program scaffold:

cbt-programs scaffold cbt-shift-work-sleep \
  --title "CBT for Shift Work Sleep" \
  --condition "Shift Work Sleep Disturbance" \
  --creator ai-assisted

5. Track progress in markdown:

cbt-programs progress cbt-shift-work-sleep \
  --title "Added module drafts" \
  --summary "Drafted first two modules and updated recommendation signals"

6. Improve an existing program:

cbt-programs improve cbt-shift-work-sleep \
  --llm \
  --llm-plan-file ./llm-improve-plan.json \
  --goal "Improve adherence and add relapse prevention support for sleep setbacks" \
  --dry-run \
  --note "Improved content quality and review status"

Or use guided mode:

cbt-programs improve cbt-shift-work-sleep --interactive

LLM mode expects the host agent to supply the plan JSON (--llm-plan or --llm-plan-file), so no separate API call is needed by this tool.

7. Build assistant prompt for any agent system:

cbt-programs agent-prompt cbt-shift-work-sleep --lang en --goal "Prepare week 1 plan"

Translation workflow

• Add a new locale file under data/programs/<id>/locales/<lang>.json
• Add translation metadata to program.json.translations[]
• Set status to one of: draft, human-reviewed, expert-verified
• Run cbt-programs validate <id>

Legacy import workflow

node scripts/import_legacy_yaml.js \
  --source /Users/birger/Community/egen_journal/backend/content/programs \
  --status draft

Requires Python + PyYAML for parsing legacy YAML.

Publishing rules

For status: published, ensure:

• At least one translation is human-reviewed or expert-verified
• lineage flags are truthful and complete
• Safety disclaimer exists
• progress/ has markdown entries

Governance details: references/governance.md