CLI-first, markdown-first Jira skill — ~80% fewer tokens than Atlassian MCP on everyday tasks.

An Agent Skill for managing Jira work items with Atlassian CLI (acli) and mdadf for Markdown-to-ADF conversion.

• Zero setup — asks for your project name on first use. No Python, no Docker, no config files required.
• Rich text that works — write Markdown, get proper Jira formatting via mdadf. No raw JSON or plain text.
• View, create, edit, search, transition, assign, and comment

Saves ~80% of Jira-related tokens on everyday tasks — more context room for actual work.

Top wins per operation:

• Listing tickets: 94% fewer tokens
• Session startup: 74% fewer
• Rich-markdown creates: 74% fewer

Agent-loop (ship a ticket end-to-end): ~82% fewer tokens per dev cycle (~5.6×) — fits ~6× more complete cycles in one 200K-token session.

Full numbers, methodology, and cost translation: benchmark/README.md.

npx skills add chenhunghan/jira-skill

Requires acli and mdadf — the skill will check for both on first run.

On first use the skill asks which Jira project to use and where to save the config — no manual setup needed. If you prefer to pre-configure:

# Global — shared across all repos (recommended for most users)
mkdir -p ~/.config/jira-skill
echo '{ "defaultProject": "YOUR_PROJECT_KEY" }' > ~/.config/jira-skill/config.json

# Per-repo — overrides the global config for this repo only
echo '{ "defaultProject": "YOUR_PROJECT_KEY" }' > .jira-skill.json

MIT