
Four steps from install to a clean PR check.
No background daemon, no telemetry pipeline to wire up, no SDK to swap. Skene reads the analytics code you already wrote.
Install
Drop Skene into your coding agent as an MCP server, or add the GitHub Action to your CI. Both run on the same engine. The MCP catches drift before the agent commits. The Action catches anything that bypassed the MCP.
Index
Skene parses the repository and finds every call into PostHog, Mixpanel, Segment, Amplitude, Heap, and Snowplow. It records the event name, the payload shape, the file, the line, and whether the call sits inside a conditional. That index is the baseline.
Compare on every PR
When a PR touches a file with analytics in it, Skene rebuilds the index for the new state and diffs it against the baseline. Removed calls, renamed events, altered payloads, and conditional changes that drop firing all land as PR comments. The comment names the event, the file, and the line.
Resolve, then merge
Each finding is either a real regression (fix it) or an intentional change (mark it accepted; the index updates). Merging without doing one or the other fails the check. The baseline tracks the codebase from then on.
See what your last 10 PRs quietly broke.
Install in a minute. The first repo audit runs in about as long as your CI does.