file-issue - SKILL.md Agent Skill

name: file-issue description: File a GitHub issue for a bug or improvement found this session.

Skill: File GitHub Issue

Create a GitHub issue in marin-community/marin from bugs, regressions, or improvements identified in the current conversation.

Background

Read first:

@AGENTS.md

Issue Kinds and Body Structure

Pick the kind, then use the matching body structure below. There are no GitHub issue templates — these structures live here.

Kind	When to use	Labels
bug	A bug or regression was found	`bug`, `agent-generated`
task	An improvement, refactor, or feature request	`agent-generated` + priority if known
experiment	An experiment needs tracking	`experiment`, `agent-generated`

Bug body

**Describe the bug**
<what is broken -- concrete symptoms, error messages>

**To Reproduce**
1. <step>
2. <step>

**Expected behavior**
<what should happen instead>

**Additional context**
<root cause analysis, file:line references, suggested fix if known>

Task body

## Description
<what needs to be done and why -- enough context for anyone on the team>

### Definition of Done
<specific, testable completion criteria>

Experiment body

## TL;DR

<One-paragraph current summary. Leave blank only when the work is just being kicked off.>

## Description

<Context someone outside the thread can understand.>

## Hypothesis or Goal

<What are you trying to learn, fix, or achieve?>

## Status

<Current state; update as evidence lands.>

## Links

* Logbook:
* W&B Report:
* Important updates:

## Decision Log

## Conclusion

Workflow

1. Gather Context from Conversation

Extract from the conversation:

What is broken or missing -- concrete symptoms, error messages, failing test output.
Where it happens -- file paths, line numbers, module names.
How to reproduce -- steps, commands, or minimal config that triggers it.
Root cause (if known).
Severity -- blocks work, causes data loss, or cosmetic?

If it's ambiguous what to file, ask the user before proceeding.

2. Classify the Issue

Pick the kind (bug, task, or experiment). If unsure, ask the user.

3. Duplicate Check

Search for existing issues first:

gh issue list --repo marin-community/marin --state open --search "<keyword>"

If a match exists, tell the user and offer to comment on it instead.

4. Draft the Issue

Title: Short imperative sentence under 80 characters, optionally prefixed with a scope tag (e.g. [levanter] Fix gradient accumulation off-by-one).

Body: Use the section structure for the chosen kind (see above).

Rules for the body:

No filler ("I noticed...", "During our conversation...").
No markdown images or tables.
Reference code with file:line links, not inline dumps.
Keep bug and task issues under ~200 words; experiment issues may be longer when the tracking context needs it.
Include error messages or stack traces in code blocks, trimmed to the relevant frames.
For task issues: include a concrete Definition of Done.
For bug issues: include numbered reproduction steps.

5. Confirm or File Directly

If the user explicitly asked to file an issue, skip the preview — file it and share the link. If the agent surfaced the issue (not explicitly requested), show the drafted title and body and wait for approval or edits.

6. File the Issue

Write the body to a uniquely named temp file, then pass it with --body-file. Do not inline the body with shell substitution (--body "$(cat <<'EOF' ...)") — multiline text can be corrupted by pasted output or escaping mistakes. Do not reuse a fixed path like /tmp/issue-body.md; concurrent agent runs can overwrite each other's drafts on shared hosts.

body_file="$(mktemp "${TMPDIR:-/tmp}/issue-body.XXXXXX.md")"
trap 'rm -f "$body_file"' EXIT

cat > "$body_file" <<'EOF'
<body>
EOF

gh issue create --repo marin-community/marin \
  --title "<title>" \
  --label "agent-generated" \
  --body-file "$body_file"

Add kind-appropriate labels (bug, experiment). If a relevant label does not exist, skip it rather than creating new labels. For task issues, add a priority label (p1, p2, p3) if the user specifies one or severity is clear.

Before creating the issue, re-open the body file and verify it contains no unrelated shell output (pre-commit logs, pytest session headers, prompt transcripts). If it does, clean the draft before posting.

7. Report Back

Print the issue URL.

Writing Style

Follow the terse style from fix-issue: every sentence conveys new information; no preamble or editorializing; no restating code a link covers; annotate code links, don't narrate them.

Tasks

Extract bug/issue details from conversation
Classify as bug, task, or experiment
Run duplicate check against open issues
Draft issue title and body using the matching kind structure
Show draft to user for confirmation
File issue with gh issue create
Report issue URL to user

Rules

Never credit yourself in the issue.
Always add the agent-generated label.
Confirm with the user before filing only when the agent surfaced the issue (not when the user explicitly asked to file).
If the conversation does not contain a clear bug or actionable improvement, say so and ask the user what they want to file.
Always use the section structure for the matching kind (see "Issue Kinds and Body Structure" above).