What problem does it solve? Manual and exploratory testing normally produces prose claims that a quality gate cannot verify. This Skill structures manual testing as a charter → session → debrief workflow (SBTM) so every command run leaves a telemetry anchor and the result is written as a machine-readable manual check that a Definition-of-Done gate can re-read and adjudicate. ## Core Features & Use Cases - SBTM workflow: Define a one-paragraph charter, run a timeboxed session of real commands, then debrief into a structured verification check. - Anchored evidence: Arm a session via HARNESS_MANUAL_TEST_SESSION=1 or a state-dir marker so the manual_test_anchor hook records each Bash command and its output as a citable anchor. - Admissibility tiers: Evidence is classified as claimed, anchored, or anchored-plus-cosign; only an anchor plus a human co-signer distinct from the author is admissible at a hard gate. - Use Case: Before merging a change to the password-reset flow, run a 30-minute exploratory session probing expired tokens, record anchored results, get a reviewer to co-sign the charter, and attach the manual check to verification.json so the hard gate passes. ## Quick Start Ask the agent to run a manual test session with the charter "verify the password-reset flow rejects an expired token" and produce anchored evidence for the verification gate.
