Codify a proven run with replay

These examples and illustrative results are independently authored teaching materials, not measured model results.

Use case

Codify an accepted read-only extraction as a tool. Teaching HTML contains id-A/id-B and titles. Failed exploratory selectors do not belong in the procedure, and object existence is insufficient.

Mechanism

Retain final successful navigation/read trace, intent and accepted input/output identity, excluding failed/unrelated exploration. Separate pure HTML parsing from browser I/O. Replay fixed snapshots with independent ID/order/field/count assertions, review permissions/dependencies and enable only within authorized installation after checks. Verify live changes separately with snapshot dates.

Bad example

Reconstruct from memory, keep every attempted selector and call an object return replayable/live-ready.

Good example

Codify only the accepted teaching sequence and parse the snapshot into id-A/id-B with original titles/order. Assert exact values and remove failed selectors; test navigation separately and review code/authority before installation. Snapshot success guarantees no future page.

Why the change matters

Evidence-based codification reduces memory errors, and pure replay supplies deterministic accepted references. Exact assertions catch wrong selectors/fields beyond shape smoke tests.

Observable expectation

Teaching replay returns two matching IDs/titles/order; empty objects, swapped order or wrong titles fail. New snapshots require explicit versions/expectations rather than weakening tests. No tool was installed/executed here.

Limits

Snapshots cover historical input only; pages/authentication/pagination/navigation change. Pure parsing does not test I/O and generated code/source remain reviewable. One accepted run does not establish all-scenario reuse.