| name | verify-before-trust |
| description | Use after a claim or change needs proof before it is trusted. Owns post-claim and post-change verification only. |
| category | core |
| source_group | hermes-operational |
| compatibility | opencode |
When To Use
Load this after a claim, fix, rewrite, migration, generated artifact, or handoff needs proof before it is trusted. The skill owns post-claim and post-change verification.
Use it when the work is already asserted as done, correct, aligned, or improved, and the next responsible move is to test that assertion against evidence.
Do Not Use
Do not use this for initial inspection, root-cause diagnosis, mode selection, decision packaging, or action-forcing. It starts after there is something specific to verify.
What To Do
- Identify the claim being tested in one sentence.
- Choose the smallest verification that actually bears on that claim: test, lint, diff, rendered output, replayed workflow, checksum, manual check, or acceptance command.
- Report the result without laundering uncertainty. Passing a narrow check proves only the narrow claim.
- If verification is blocked, state the blocker and the next evidence needed.
- Do not substitute confidence, consistency, or intention for proof.
Output Shape
Return the claim, verification performed, result, and any remaining unverified surface.
