cellagent-annotation

<!--
# COPYRIGHT NOTICE
# This file is part of the "Universal Biomedical Skills" project.
# Copyright (c) 2026 MD BABU MIA, PhD <md.babu.mia@mssm.edu>
# All Rights Reserved.
#
# This code is proprietary and confidential.
# Unauthorized copying of this file, via any medium is strictly prohibited.
#
# Provenance: Authenticated by MD BABU MIA

-->

---
name: cellagent-annotation
description: Cell tagger
keywords:
  - single-cell
  - markers
  - annotation
  - confidence
  - tissue
measurable_outcome: Label every provided cluster with a cell type + confidence + marker evidence (or "ambiguous") within 15 minutes per dataset.
license: MIT
metadata:
  author: CellAgent Team
  version: "1.0.0"
compatibility:
  - system: Python 3.9+
allowed-tools:
  - run_shell_command
  - read_file
---

# CellAgent Annotation

Use CellTypeAgent to interpret marker genes, annotate scRNA-seq clusters, and coordinate multi-agent workflows for downstream analysis.

## When to Use
- Automated annotation of scRNA-seq datasets without manual curation.
- Multi-step workflows (QC → clustering → annotation → DE analysis).
- Integrating multiple batches requiring consistent labeling.

## Core Capabilities
1. **Planning:** Multi-agent planner decomposes analysis goals into steps.
2. **Tool execution:** Generates Scanpy/Seurat code and runs it autonomously.
3. **Self-correction:** Detects execution errors and retries with fixes.

## Workflow
1. Gather marker lists per cluster, plus species/tissue context and optional atlas references.
2. Run CellTypeAgent (`pip install -r requirements.txt` then `python repo/main.py --data data.h5ad --goal annotate`).
3. Review outputs for supporting markers; downgrade ambiguous clusters when signals conflict.
4. Produce final table (cluster, label, confidence, supporting markers, notes) and cite references when used.

## Example Usage
```bash
python3 Skills/Genomics/Single_Cell/CellAgent/repo/main.py --data "./data.h5ad" --goal "annotate"
```

## Guardrails
- Avoid over-specific lineages if markers overlap; default to broader types.
- Flag clusters showing multiple signatures for manual review.
- Respect species/tissue differences when interpreting markers.

## References
- README + upstream paper (Mao et al., 2025 / arXiv 2407.09811).


<!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->