File a GitHub issue
Before drafting, read AGENTS.md and:
.agents/skills/writing-style/SKILL.md.agents/skills/writing-style/issues.md.agents/skills/writing-style/ai-writing-donts.md
Issue Kinds and Body Structure
Pick the smallest matching structure.
| 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
<what is broken and its impact -- concrete symptoms or error messages>
Reproduce:
1. <step>
2. <step>
Expected: <what should happen instead>
<optional: concise evidence or confirmed root cause>
Task body
<what needs to be done and why -- enough context for anyone on the team>
Done when:
<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 and classify
Extract the symptom or desired outcome, impact, location, reproduction, known cause, and severity. Ask when the issue or its kind is ambiguous.
2. Check duplicates
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.
3. Draft
Title: At most 80 characters, optionally prefixed with a scope tag. State a
factual symptom for a bug (e.g. [levanter] Gradient accumulation drops the last microbatch) and an imperative outcome for a task (e.g. [levanter] Handle partial accumulation steps). Do not add bug:, task:, or another type
prefix.
Use the body structure above. Keep the facts needed to act; link code with
file:line, trim errors to relevant frames, and omit filler, repeated titles,
diff inventories, and unnecessary implementation narration. Bugs need numbered
reproduction steps; tasks need testable completion criteria.
4. Inspect the payload
Apply the writing-style final compression pass to the exact title and body that will be sent to GitHub. For a bug or task, verify the title is at most 80 characters. Every remaining sentence must add a symptom, impact, reproduction step, observation, expected behavior, or completion criterion.
5. Apply the approval boundary
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 > "$body_file" <<'EOF'
<body>
EOF
issue_url="$(gh issue create --repo marin-community/marin \
--title "<title>" \
--label "agent-generated" \
--body-file "$body_file")"
Add kind-appropriate labels (bug, experiment). If a relevant label does not
exist, skip it rather than creating new labels. For task issues, add a priority
label (p1, p2, p3) if the user specifies one or severity is clear.
Before creating the issue, re-open the body file and verify it contains no unrelated shell output (pre-commit logs, pytest session headers, prompt transcripts). If it does, clean the draft before posting.
After creating the issue, fetch its published text with
gh issue view "$issue_url" --json title,body and correct any text added or
altered by the publishing tool.
7. Report
Print the issue URL.