Feature Description
Browser recon now uses a Playwright-style contract, while reusable adapters still use Webcmd’s existing IPage/adapter APIs. We should preserve that boundary and avoid introducing a second adapter contract.
Proposal:
- Keep
browser run as exploratory/ad-hoc Playwright-style code.
- Keep adapters on the existing stable API:
browser:false -> func(args), browser:true -> func(page,args).
- Add structured
adapterHints to recon/analyze output with:
- recommended strategy
- selector evidence
- network evidence
- state hazards
- adapter-compatible path
- explicit “do not copy Playwright code into adapter” guidance
- Only extend
IPage with narrow primitives when evals prove a missing capability.
Reason:
Letting adapters accept both Playwright-style page and IPage would increase cloud/local parity risk, repair cost, and agent confusion. A translation artifact gives agents the bridge without changing the adapter API.
Use Case
For users creating new adapters, it must get made efficiently.
Proposed Solution
No response
Alternatives Considered
No response
Feature Description
Browser recon now uses a Playwright-style contract, while reusable adapters still use Webcmd’s existing
IPage/adapter APIs. We should preserve that boundary and avoid introducing a second adapter contract.Proposal:
browser runas exploratory/ad-hoc Playwright-style code.browser:false -> func(args),browser:true -> func(page,args).adapterHintsto recon/analyze output with:IPagewith narrow primitives when evals prove a missing capability.Reason:
Letting adapters accept both Playwright-style
pageandIPagewould increase cloud/local parity risk, repair cost, and agent confusion. A translation artifact gives agents the bridge without changing the adapter API.Use Case
For users creating new adapters, it must get made efficiently.
Proposed Solution
No response
Alternatives Considered
No response