Quick Start¶
A first /oss walkthrough — from command to reviewed draft.
Before You Begin¶
Complete Installation and restart your Claude Code session so the guardrail hook and MCP server are active.
Walk Through: Finding a Good First Issue¶
This walkthrough uses the find scenario to locate a beginner-friendly issue in apache/airflow.
Step 1 — Run /oss¶
In Claude Code, type:
The /oss command reads your intent, invokes the oss-scenario-routing skill, and classifies it as the find scenario.
Step 2 — Answer the Intake Questions¶
The issue-matching skill asks up to 5 short questions before fetching data:
1. What are you hoping to achieve? (bug fix, feature, docs, learning, …)
2. What languages/frameworks are you comfortable with?
3. Is this your first contribution to apache/airflow? (yes/no)
4. Issues only, PRs only, or both?
5. Any other target projects to include?
Answer each question. The more specific your answers, the better the ranking.
Step 3 — Watch the Research¶
The oss-researcher subagent fetches live data from GitHub (via MCP or gh CLI). You will see it paginate through open issues labelled good first issue and help wanted, filtering out locked, assigned, or closed items.
Step 4 — Review the Saved Report File¶
The orchestrator saves the ranked shortlist to a file in .oss-drafts/:
Saved to .oss-drafts/find-apache-airflow-find-20260611T143000Z.md.
Review and edit this report file as needed before using it.
The file contains a ranked list of up to 10 issues with rationale. A brief summary is shown in chat:
## #1. Issue #42345: Add docstring examples to BaseSensor.poke()
- Link: https://github.com/apache/airflow/issues/42345
- Type: Issue
- Labels: good first issue, area:docs
- Last updated: 2026-06-07
- Why it fits: Docs task; matches your stated goal of learning and Python comfort.
Open .oss-drafts/find-apache-airflow-find-20260611T143000Z.md in your editor to read the full
list, add notes, or highlight the issues you want to investigate.
Step 5 — Continue with a Draft Scenario (Optional)¶
Once you pick an issue, you can use a draft scenario:
The claim scenario dispatches oss-claim-analyst, which checks assignees, linked PRs, and recent intent comments, then returns a verdict. If the issue is free, smart-questions drafts a claim comment. The orchestrator saves it to a file and shows:
DRAFT — saved to
.oss-drafts/claim-apache-airflow-issue42345-20260611T143200Z.md. Review and edit this file before sending. This plugin will NOT post, comment, push, or send anything. Sending is handled by the separate execution/submission plugin.
The draft text in chat will look like:
Hi, I'm [your name], a Python developer interested in contributing to Airflow.
I'd like to work on issue #42345 (Add docstring examples to BaseSensor.poke()).
Is there anything specific you'd like me to keep in mind? Happy to discuss the approach first.
Open the saved file, edit it to your liking, then use the execution/submission plugin to post it.
What the Plugin Will Never Do¶
- Post a comment on GitHub.
- Create or close an issue or PR.
- Push a branch.
- Send any outbound request with mutation intent.
All such attempts are blocked by the PreToolUse guardrail hook — see Draft-Only Guardrail.
Back to docs index | Installation | Next: Using /oss