What problem does it solve? Rolling back a broken production deploy is riskier than it looks: on stateful platforms like Odoo, Django, or Rails, reverting code does not uninstall modules, undo migrations, or remove user-facing artifacts, and a stale revert PR can silently rip out newer work. This Skill turns "revert that commit" into a correct, honestly-described rollback procedure. ## Core Features & Use Cases - Precise culprit identification: Uses git log -S, --since, and --stat to bind a symptom to one introducing commit, distinguishing causes from mere exposers of pre-existing behavior. - Semantic dependency checking: Detects later commits that build on the culprit's symbols or files, so a plain revert does not break downstream code. - Honest revert PRs: Structures the PR body to state what rolls back, what merging does operationally, what the revert does NOT undo (installed modules, migrations, sent emails, cached assets), and the re-landing path. - Decay rule for stale reverts: Re-validates or closes revert PRs that have sat open while the base branch moved, treating closure as a valid outcome. - Use Case: Yesterday's deploy broke a customer-facing page on your Odoo instance. Use this Skill to find the culprit commit, verify nothing later depends on it, syntax-check the reverted tree, and write a revert PR that explicitly lists what the rollback will not undo. ## Quick Start Ask the agent to revert the commit that broke yesterday's deploy and write a rollback PR that states what the revert does not undo.
