Follow-up after #14 and #20.
Context
direct-cli now has a stable integration_write suite with replayable VCR cassettes.
The current contract is effectively 10 passed, 9 intentionally skipped because some
write scenarios are limited by Yandex Direct sandbox behavior rather than by missing CLI implementation.
This issue tracks the next phase: deciding how to expand write-test confidence beyond the
current sandbox ceiling.
Problem
Some write flows still cannot be validated as successful end-to-end in the current Yandex Direct sandbox.
This leaves a ceiling on how much live integration confidence direct-cli can gain from sandbox replay/record alone.
Roadmap
1. Classify every currently skipped write scenario
For each currently skipped scenario, decide whether it is:
- permanently sandbox-limited
- fixable with better fixtures or resource chains
- only testable outside sandbox
- better covered by stronger dry-run or cassette regression guards
2. Re-evaluate specialized fixtures for nested resources
Check whether better fixture topology can unlock more reliable success-path coverage for:
ads
keywords
keywordbids
audiencetargets
dynamicads
smartadtargets
3. Decide whether a separate non-sandbox integration tier is needed
If some scenarios cannot be made reliable in sandbox, decide whether they should move into a separate integration tier with different prerequisites and expectations.
4. Document the coverage matrix explicitly
Make covered vs sandbox-limited vs other-tier-required explicit in repo docs and/or test headers so issue closure is based on a stable contract rather than on inference.
Acceptance Criteria
- every currently skipped write scenario has an explicit classification
- each classification has a chosen next-step testing strategy
- repo documentation clearly distinguishes:
- replay-covered success-path tests
- intentional sandbox-limited skips
- scenarios that require a different testing tier
Notes
This is not a bug issue. It is a test-strategy roadmap issue for going beyond the current Yandex sandbox ceiling.
Follow-up after #14 and #20.
Context
direct-clinow has a stableintegration_writesuite with replayable VCR cassettes.The current contract is effectively
10 passed, 9 intentionally skippedbecause somewrite scenarios are limited by Yandex Direct sandbox behavior rather than by missing CLI implementation.
This issue tracks the next phase: deciding how to expand write-test confidence beyond the
current sandbox ceiling.
Problem
Some write flows still cannot be validated as successful end-to-end in the current Yandex Direct sandbox.
This leaves a ceiling on how much live integration confidence
direct-clican gain from sandbox replay/record alone.Roadmap
1. Classify every currently skipped write scenario
For each currently skipped scenario, decide whether it is:
2. Re-evaluate specialized fixtures for nested resources
Check whether better fixture topology can unlock more reliable success-path coverage for:
adskeywordskeywordbidsaudiencetargetsdynamicadssmartadtargets3. Decide whether a separate non-sandbox integration tier is needed
If some scenarios cannot be made reliable in sandbox, decide whether they should move into a separate integration tier with different prerequisites and expectations.
4. Document the coverage matrix explicitly
Make
coveredvssandbox-limitedvsother-tier-requiredexplicit in repo docs and/or test headers so issue closure is based on a stable contract rather than on inference.Acceptance Criteria
Notes
This is not a bug issue. It is a test-strategy roadmap issue for going beyond the current Yandex sandbox ceiling.