issue-bot

https://github.com/johnalison/issue-bot/tree/master

Logs

11 February 2026 Wednesday

  • [X] Idea for \test function… push a branch and monitor the gitlab CI status
  • Now run-local-CI not working on LPC… removing and reinstalling pixi …
  • Got the CI running on falcon… will try to setup there

10 March 2026 Tuesday

  • Trying to run the CI on falcon… failing with 

    WorkflowError:
The apptainer or singularity command has to be available in order to use apptainer/singularity integration.
  • Waiting for AE to fix…

09 March 2026 Monday

  • Do we need to pass context that it is being invoked headlessly by the issue-bot ?

Prompt for /plan-fix

I'm building an automated GitLab issue bot that invokes Claude Code headlessly to fix issues. I need you to create a `/plan-fix` slash command for this repo.

The skill lives at `.claude/commands/plan-fix.md`. Create it now.

The skill will be invoked by the bot (and optionally by humans) before making any code changes. When run, it should instruct Claude to:

  1. Read the current issue context from the prompt (title, description, URL will already be in context)
  2. Explore the codebase to understand what needs to change — use the module map in CLAUDE.md as a starting point
  3. Produce a structured plan with these sections:
    • Affected files — list every file that will need to change and why
    • Proposed changes — for each file, a brief description of what will be added/modified/removed
    • Assumptions & judgement calls — anything ambiguous in the issue and how you plan to resolve it
    • Risk — anything that could break existing behavior
  4. Make NO code edits during this step

After creating the file, invoke `/plan-fix` yourself on this example issue to test it: "bot:fix: the trigger emulator is not correctly handling 2023 run conditions"

Show me the output and we can iterate on the skill definition together.

Prompt for /test-local-CI

I'm building an automated GitLab issue bot that invokes Claude Code headlessly to fix issues. I need you to create a `/test-local-CI` slash command for this repo.

The skill lives at `.claude/commands/test-local-CI.md`. Create it now.

The skill will be invoked after code changes have been made, to validate the fix before the bot commits and opens an MR. When run, it should instruct Claude to:

  1. Copy the grid proxy from the user's home directory: cp ~/x509up* proxy/x509proxy If this fails, warn that the proxy may be missing or expired and stop.

  2. Run CI tests from the barista root using: source coffea4bees/scripts/run-local-ci.sh <job>

    If $ARGUMENTS are provided, run only those specific jobs. If no $ARGUMENTS, default to the fast code tests:

    • codeanalysishelpers
    • codejetclustering
    • codetrigemulator
  3. Check output logs in CIoutput/ for FAILED or ERROR lines.
  4. Report a clear summary: which tests passed, which failed, and any relevant log excerpts for failures.

After creating the file, look at the available CI jobs in coffea4bees/workflows/SnakefiletestCI.smk and make sure the skill's documentation lists them so Claude knows what's available when choosing which tests to run.

Then invoke `/test-local-CI` to verify it works. Show me the output so we can iterate on the skill definition together.

Backlinks