agentleFS
ClickHouseClickHouse/loom/.claude/skills/loom-file-issue/SKILL.md

loom-file-issue

Skill87 starsChanged 3 days agoApache-2.0View on GitHub

Files or updates a Loom GitHub issue — duplicate check, a reproduction with versions, the matching issue form and label, a redacted body. Use when asked to open, file or report an issue, bug, regression or feature request, or to add findings to one. Not for security reports (SECURITY.md).

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Keep the error.code and the X-Request-Id of the failing response.

    line 26
  2. Distinctive:

    Redaction: remove tokens, keys, JWTs, passwords, customer data, host names, internal paths and namespaces from logs before pasting, then, …

    line 51
  3. Very specific:

    Write in the voice of loom-writing and run its checker on the body file.

    line 52

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Filing a Loom issue

An issue is worth filing when a maintainer can act on it without asking a question back. Every issue goes through one of the forms in .github/ISSUE_TEMPLATE/ (blank issues are off).

1. Is it already tracked?

gh issue list --repo ClickHouse/loom --state all --search "<two or three distinctive terms>" -L 20
gh pr list --repo ClickHouse/loom --state all --search "<same terms>" -L 10

With the Loom plugin connected, also call its loom_tracker_duplicate_check tool with the title, body and any stack frames: it matches by meaning, not phrase. A duplicate gets a comment with your new evidence on the existing issue, not a second issue. A fix already on main gets no issue; say which commit fixed it.

2. Reproduce it

  • Reproduce on the latest release or main, on a private server (loom-dev-setup), before writing. If you could not reproduce it, say so: in "How this was found" where the form has it, otherwise in "Additional context", or in the main field of a form with neither.
  • Shrink to the smallest sequence: the ops or SDK calls with their inputs, the non-default LOOM_* settings, whether auth is on, the rows it depends on and the latest applied migration (SELECT max(filename) FROM loom._migrations).
  • Keep the error.code and the X-Request-Id of the failing response; loom-debug follows that id to the cause.
  • A regression names the PR that caused it (loom_code_bisect_candidates ranks the merged PRs of a good/bad window when the plugin is connected): Caused by: <full PR URL>.

3. Pick the form

The label column is a label that exists in the repository; a form's own labels: that does not exist is dropped silently.

Form (.github/ISSUE_TEMPLATE/)LabelFor
30_bug-report.yamlbugan op, SDK call, CLI command, worker, migration or hook does the wrong thing
40_recall-quality.yamlbugsearch, ask or a context pack returned a wrong result, missed a memory, or failed to abstain
50_performance-issue.yamlbugslower, or reads more from ClickHouse, than expected
60_upgrade-compatibility.yamlbugan upgrade broke a client, a migration, a field or a setting
70_installation-issue.yamlbugdoes not install, build, boot or connect
20_feature-request.yamlenhancementa workflow Loom should support, starting from the use case
80_documentation-issue.yamldocumentationa page is wrong, missing or unclear
10_question.yamlquestionhow to do something

4. Write the body

gh issue create cannot fill a YAML form, so write the body with one ### <label> heading per form field, in the form's order, using the labels exactly as the form shows them (read the YAML). Fill every required field; drop optional ones you have nothing for.

  • Title: the wrong behaviour as a fact, under ~80 characters. "memory.search drops pinned rows when limit is below the pin count", not "Search bug" or "Issue with pinned memories".
  • Environment: paste loom report (it prints versions and loom doctor, never secret values).
  • How this was found: pick the option that is true. An agent that found the problem and has not had a person reproduce it says so; that tells a maintainer how much to trust the reproduction.
  • Redaction: remove tokens, keys, JWTs, passwords, customer data, host names, internal paths and namespaces from logs before pasting, then, in a form with a Redaction field (bug, recall quality, performance, upgrade, installation), end the body with its required checkbox: ### Redaction and - [x] I removed tokens, keys, passwords and personal or customer data from everything above.
  • Write in the voice of loom-writing and run its checker on the body file.
python3 .claude/skills/loom-writing/scripts/check_prose.py /path/to/issue.md
gh issue create --repo ClickHouse/loom --title "<title>" --label "<label from the table>" --body-file /path/to/issue.md   # without triage rights: drop --label

Creating an issue is public: show the title and body to the requester and file only after they agree, unless they asked you to file it outright.

5. After filing

  • Link it where it belongs: Closes: <issue URL> in the PR that fixes it, Related: from other issues.
  • New evidence goes in a comment on the issue (gh issue comment <n> --body-file …), not an edit that rewrites what others have already read.
  • Close only issues you opened, and only with the reason: the fixing PR, or why it is not a bug.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.