9

Source document

[2026-03-28 23:30] traves_theberge: Option 1: GitHub Action (CI) PR created by Orchestra → GitHub Action triggers openlens → Finds issues → posts PR review comments → Human or Orchestra creates issues from comments → Circle continues Option 2: Git hooks (in the agent's worktree) Agent commits code → pre-push hook runs openlens → Blocks push if critical findings → Agent sees the error, fixes it → All happens inside the agent's session Option 3: Platform hooks (in the agent CLI) Claude/Codex/Gemini hooks config → PostToolUse hook on git commit/push → Runs openlens inline → Agent gets findings in its context → Fixes before finishing Option 4: Manual (user runs it) User reviews PR → Runs: openlens run --branch main → Creates issues from findings [2026-03-28 23:31] traves_theberge: do i add complexity of do i keep it barebones for the user to configure.. i like the bare bones and have the user do it on their own ci/ci pipelines. [2026-03-28 23:35] traves_theberge: fuck it. users can trigger it on PR or configure openlens on the agents. [2026-03-28 23:35] traves_theberge: not my problem [2026-03-28 23:35] traves_theberge: https://tenor.com/view/not-my-circus-not-my-monkey-gif-11313350 [2026-03-28 23:36] traves_theberge: but maybe i could add a PR management, and CI/CD management, and Agent management (hooks, Skills, plugins, instructions, mcp, et.) for projects. [2026-03-28 23:37] traves_theberge: seems resonable [2026-03-28 23:38] ajaxdavis: i think first question to answer is, does openlens replace my traditional lint/checks, if i buy into openlens should i use it solely [2026-03-28 23:38] ajaxdavis: if i don't, there is a race to the source of truth between them as qualitive measures on the code base [2026-03-28 23:40] traves_theberge: there is a line of reasoning here i like.. Somthing like Agent does some work(claude), another agent checks the work(openlens), another agent tests the work(Claude + Playright + test plans), feedback loop to user / issue creation [2026-03-28 23:41] traves_theberge: (files: image.png) [2026-03-28 23:41] ajaxdavis: my vote is option a) as the standard for the project if i want to do it locally in my own workflow, i can. [2026-03-28 23:45] traves_theberge: the issue with the tools like https://github.com/millionco/expect it only works for one side of the testing ( which is like SASS ) but testing for backend needs its own testing method. [2026-03-28 23:45] traves_theberge: anyway thinking too far ahead at the moment [2026-03-28 23:48] traves_theberge: i want to balance with local and remote workflows, i think imo, linting/checks localy may differ from the pipelines in CI/CD [2026-03-29 00:03] ajaxdavis: can i only move something to ToDo if it has an assignee? (files: Screenshot_from_2026-03-29_10-03-27.png) [2026-03-29 00:03] ajaxdavis: doesnt seem to let me drop it