elder-care-casework

# Project guidance for Copilot agents

This repository contains reusable, privacy-first elder-care casework workflows. Keep populated case workspaces, raw records, recordings, transcripts, generated findings, correspondence drafts, and prompt/log artifacts out of git.

## Customization layout

- Canonical Cowork / Claude Desktop plugin skills live under `elder-care-casework/skills/`.
- VS Code and GitHub Copilot CLI agents live under `.github/agents/`.
- VS Code skill wrappers live under `.github/skills/` and point back to the canonical plugin skill files.
- VS Code prompt files live under `.github/prompts/` for quick task starts.

## Privacy and evidence rules

- Default to plan-only for offload workflows. Do not launch Copilot CLI, local model, transcription, diarization, or image-analysis jobs unless the user explicitly asks to run them.
- Keep PHI in private local output folders. Chat summaries should use counts, categories, and non-identifying filenames unless the user explicitly requests quoted material.
- Preserve citations exactly: source filename, page or timestamp, speaker if known, and confidence. Mark uncertainty instead of guessing.
- Separate factual evidence triage from medical or legal advice. Skills may surface candidates for human review; they do not diagnose, interpret law authoritatively, or replace counsel.

## Workflow-specific entry points

- For SNF PDF records, load `elder-care-casework/skills/snf-records-review/SKILL.md`, `RUBRIC.md`, and `REVIEWER_PROMPT.md` before reviewing.
- For audio recordings, transcripts, plan-of-care meeting audio, voicemail, or facility phone calls, load `elder-care-casework/skills/audio-transcript-review/SKILL.md` and `TRANSCRIPT_REVIEW_PROMPT.md`.
- For expensive local analysis with Qwen / GLM / Gemma-class models, LM Studio/`lms`, LM Studio OpenAI-compatible API, Goose CLI, Ollama, llama.cpp, vLLM, ffmpeg, Whisper, Parakeet, OCR, or vision models, load `elder-care-casework/skills/local-analysis-offload/SKILL.md` and keep the run bounded and auditable.

<!-- mermaid-ai-skills:start -->
## Mermaid Diagrams

When the user asks to create, edit, or visualize a diagram, follow the
instructions in `.github/instructions/mermaid.instructions.md`.
<!-- mermaid-ai-skills:end -->