Assess the risk of the current diff and exit non-zero when human review is required: tessl change risk, with tessl change risk init writing starter configuration (including a “conservative starter” preset).
The exit-code behavior is what makes it interesting rather than advisory. This is a machine-readable answer to does a person need to look at this?, which is the routing decision that matters most once agents open more PRs than anyone can read. It converts reviewer attention from a constant into something allocated by measured blast radius — pattern-scale-adaptive-planning applied to the review budget rather than to planning ceremony.
Its nearest counterpart is pstack-blast-radius, and the contrast is instructive. Both answer how dangerous is this change? — but pstack’s is a workflow a developer runs that demands one safety fact be proven by executing code, and Tessl’s is a gate in CI that emits a verdict and an exit code. Depth versus throughput, and the split follows the layer: a framework optimizes one engineer’s rigor, a platform optimizes an organization’s flow.
Tessl pairs it with tessl-change-verify: risk decides whether to look, verify decides what fails.
See Also
- tessl-change-verify — the paired verifier pass over changed files.
- tessl-code-review — the full PR review this triages into.
- pstack-blast-radius — the framework-side counterpart, proof-driven rather than gate-driven.
- stage-review — the canonical stage this implements.