Artifact Management
Every task must have an artifact.json that declares all output files.
Setup
mkdir -p "tasks/${CLAUDE_SESSION_ID}/reports"
mkdir -p "tasks/${CLAUDE_SESSION_ID}/notes"
cat > "tasks/${CLAUDE_SESSION_ID}/artifact.json" << 'EOF'
{
"artifacts": []
}
EOF
Adding Artifacts
Read the current file, append new entries, and write it back. Each entry needs:
| Field | Required | Description |
|---|---|---|
id |
Yes | Unique identifier (e.g. research-report) |
path |
Yes | File path relative to /bunny-agent |
mimeType |
Yes | MIME type of the file |
description |
No | Brief description of the content |
Example
{
"artifacts": [
{
"id": "research-report",
"path": "tasks/${CLAUDE_SESSION_ID}/reports/report.md",
"mimeType": "text/markdown",
"description": "Main research report with findings and analysis"
},
{
"id": "source-notes",
"path": "tasks/${CLAUDE_SESSION_ID}/notes/sources.md",
"mimeType": "text/markdown",
"description": "Research sources and citations"
}
]
}
Common MIME Types
| Type | Extension |
|---|---|
text/markdown |
.md |
text/plain |
.txt |
text/html |
.html |
text/csv |
.csv |
application/json |
.json |
application/pdf |
Rules
- Always call this skill BEFORE starting research
- Use
${CLAUDE_SESSION_ID}in paths — it auto-resolves in this file - Update
artifact.jsonwhenever you create a new output file - Every task must produce at least one artifact (typically a Markdown report)