Verification and evidence
Only the verifier says done. Everything else is a claim.
A step that ran is not a step that worked. Exit codes, ‘success’ pages, screenshots and a model’s confidence carry no authority. When the work is finished, an independent check re-reads what should now be true and compares it with what is.
What evidence looks like
| Kind | Example | Recorded |
|---|---|---|
| command | the command that was run and its exit | output, exit code, duration |
| test | a test run | what ran, what passed, what failed |
| diff | a change to files | the patch |
| screenshot | what a screen showed | the image |
| source | a page, document or message that was read | reference and excerpt |
Every evidence item has a status — pass, fail or unknown — and a confidence. Anything that could not be checked is listed as unknown rather than assumed.
Repair
When a check fails, PHRONEN fixes the cause and re-checks the original outcome — up to the task’s repair limit (three by default). Then it stops and reports what holds and what does not.
Was this page helpful?