Contributing

We develop Infrahub for customers and with the community. We are open to pull requests. You can also discuss your intentions via

Contribute to Infrahub

Please see our Development Docs for a guide to getting started developing for Infrahub.

We maintain a list of issues that are appropriate to newcomers to the Infrahub development community. This is a great place to get started!

Bug pipeline slash commands

Infrahub ships an AI-assisted bug-fixing pipeline built on GitHub Agentic Workflows. Contributors and maintainers drive it from GitHub by posting slash commands on issues and pull requests. Each command triggers one agent; agents pass state to each other through markers (AGENT_ANALYSIS_COMPLETE, AGENT_TEST_COMPLETE, AGENT_FIX_COMPLETE, AGENT_REVIEW_VERDICT: …) embedded in issue comments and PR bodies. The reviewer is not invoked via a slash command — it runs automatically when a pipeline PR is opened or updated.

Pipeline flow:

issue: /bug-analyze ──► analyst posts root cause (AGENT_ANALYSIS_COMPLETE)
issue: /bug-tdd     ──► test-writer opens draft PR (AGENT_TEST_COMPLETE)
PR opened/synced    ──► reviewer auto-runs → TEST_APPROVED | TEST_CHANGES_REQUESTED
PR:    /bug-fix     ──► fixer pushes fix (AGENT_FIX_COMPLETE)
PR synced           ──► reviewer auto-runs → FIX_APPROVED | FIX_CHANGES_REQUESTED

The pipeline uses a single PR throughout: /bug-tdd opens it, /bug-fix adds the fix to the same branch. To revise a rejected step, re-run the same command on that PR — /bug-tdd after TEST_CHANGES_REQUESTED, /bug-fix after FIX_CHANGES_REQUESTED.

Running the pipeline on community-reported issues

Issues from outside contributors must be approved by an OpsMill maintainer before the pipeline can run on them. A maintainer applies the state/ai-pipeline-ready label after reviewing the report; once labeled, anyone can post /bug-analyze. The label must stay on for the duration of the pipeline. Maintainer-authored issues skip this step.

/bug-analyze

• Where to invoke: comment on the bug issue.
• What it does: reads the issue, explores the codebase, posts a root-cause analysis and fix strategy as a comment on the issue, ending with AGENT_ANALYSIS_COMPLETE. Does not write code or open a PR.
• Preconditions: none beyond a bug issue with enough detail to investigate. If the issue is unclear, the agent labels it state/need-more-info and stops without the completion marker.
• Owning workflow: .github/workflows/bug-agent-analyst.md

/bug-tdd

• Where to invoke:
  • On the bug issue (initial test mode) — opens a draft PR with a failing test.
  • On the resulting test PR (revision mode) — reworks the test based on reviewer feedback.
• What it does: writes a single failing test that reproduces the bug, verifies it fails for the right reason, opens (or updates) a draft PR against stable from a branch named ai-bug-pipeline-<issue_number>-<slug>, and stamps AGENT_TEST_COMPLETE in the PR body. The branch prefix is what the reviewer keys off — do not rename the branch. Production code is never modified.
• Preconditions:
  • Initial mode: a comment from the bug pipeline bot on the issue containing AGENT_ANALYSIS_COMPLETE.
  • Revision mode: PR body contains AGENT_TEST_COMPLETE, does not contain AGENT_FIX_COMPLETE, and a prior reviewer comment with AGENT_REVIEW_VERDICT: TEST_CHANGES_REQUESTED exists.
• Owning workflow: .github/workflows/bug-agent-test.md

/bug-fix

• Where to invoke: comment on the pipeline PR opened by /bug-tdd.
• What it does: implements the fix on the PR branch, verifies the replication test now passes, runs format/lint/codegen/unit tests, adds a changelog fragment, updates the PR title/body from the project template, and stamps AGENT_FIX_COMPLETE. In revision mode it applies the reviewer's requested changes.
• Preconditions:
  • Initial mode: PR body contains AGENT_TEST_COMPLETE and a prior reviewer comment with AGENT_REVIEW_VERDICT: TEST_APPROVED. Must not already contain AGENT_FIX_COMPLETE.
  • Revision mode: PR body contains AGENT_FIX_COMPLETE and a prior reviewer comment with AGENT_REVIEW_VERDICT: FIX_CHANGES_REQUESTED.
• Owning workflow: .github/workflows/bug-agent-fix.md

Bug reviewer (automatic)

• Where to invoke: not invoked manually — runs automatically on pull_request events (opened, synchronize, edited, reopened) for PRs whose head branch starts with ai-bug-pipeline-.
• What it does: picks test-review or fix-review mode based on the PR body markers, posts a verdict comment whose first line is one of AGENT_REVIEW_VERDICT: TEST_APPROVED | TEST_CHANGES_REQUESTED | FIX_APPROVED | FIX_CHANGES_REQUESTED. The verdict is the gate the downstream /bug-tdd or /bug-fix run reads.
• Preconditions: PR head branch matches ai-bug-pipeline-* and the PR body contains either AGENT_TEST_COMPLETE or AGENT_FIX_COMPLETE. The reviewer self-skips if the current mode is already approved or after three iterations of the same mode (labels the PR state/needs-human-fix).
• Owning workflow: .github/workflows/bug-agent-review.md

Code of Conduct

View our CODE_OF_CONDUCT policy to find the latest information.