| name | kumo-issue |
| description | File a GitHub issue on the kumo-coding-agent repo to report a gap, bug, or feature request for the Kumo data science agent (context docs, skills, or verticals). Use when someone says "report a problem with the agent", "the agent got something wrong", "file an issue", or "request a new skill". |
| argument-hint | [description of the issue] |
| allowed-tools | ["Bash","Read","Glob","Grep"] |
File a kumo-coding-agent Issue
Create a structured GitHub issue on kumo-ai/kumo-coding-agent for agent
gaps, bugs, or feature requests.
This command works in both Claude Code (interactive) and Codex (headless).
In Codex, provide a description as the argument — interactive prompting is
not available.
User input: $ARGUMENTS
Instructions
Step 1: Validate GitHub CLI Auth
Run:
gh auth status
If gh is not authenticated, tell the user to run gh auth login -h github.com
first and stop before gathering more information.
Step 2: Gather Information
If $ARGUMENTS is provided, parse it to understand the issue.
If $ARGUMENTS is empty, ask the user these questions (skip in headless mode):
- What happened? (What were you trying to do?)
- Which skill or context doc was involved? (If known)
- What did you expect vs what happened?
Step 3: Classify the Issue
Determine the type based on the description:
| Type | Signal | Label |
|---|
| gap | "agent didn't know", "missing info", "couldn't answer" | gap |
| bug | "wrong info", "incorrect", "outdated", "says X but should be Y" | bug |
| feature | "add support for", "new skill", "new vertical", "would be nice" | enhancement |
Step 4: Identify Affected Files
Search for files related to the issue:
grep -rl "<relevant keyword>" context/ skills/
If the user mentioned a specific file, verify it exists and note the
relevant section.
Also check if this is already tracked:
grep -i "<keyword>" context/_gaps.yaml
If already tracked, tell the user and ask if they still want to file an issue.
Step 5: Compose the Issue
Title format: kumo: <concise description>
Examples:
kumo: missing Databricks Unity Catalog connector docskumo: pql-syntax.md lists MODE as valid aggregationkumo: add healthcare/clinical trials vertical
Body template:
## Type
<gap | bug | feature>
## Affected File(s)
- `<path>` (line ~N)
## Description
<Clear description of the issue>
## Expected Behavior
<What should happen / what info should be there>
## Current Behavior
<What actually happens / what the doc currently says>
## Suggested Fix
<If obvious — otherwise "Needs investigation">
## Context
<Any additional context: SDK version, customer use case, etc.>
---
*Filed via `/kumo-issue` by @REPORTER*
Step 6: Create the Issue
First, capture the reporter's GitHub identity:
gh_user=$(gh api user --jq '.login')
Replace @REPORTER in the body template with @$gh_user.
Run:
gh issue create \
--repo kumo-ai/kumo-coding-agent \
--title "<title>" \
--label "<type-label>" \
--assignee "manushmurali-kumo" \
--body "<body>"
Use a heredoc for the body to preserve formatting.
If gh is not authenticated, tell the user to run gh auth login -h github.com first
and provide the exact command to retry.
Step 7: Follow Up
After the issue is created:
- Print the issue URL
- Tell the user: "Your issue has been filed. A Kumo team member will review it and follow up with you."
- If the type is gap, suggest also adding an entry to
context/_gaps.yaml with status: open
- If the type is bug and the fix is obvious, suggest running
/kumo-pr to fix it directly
