docs(debugging): define dap failure benchmark

[627150795]

Problem

Issue #168 needs evidence about whether on-demand DAP improves diagnosis over logs and tests, but a prompt-only approximation or a debugger self-report would not be a valid comparison.

Value

Define a failure-class benchmark that measures diagnosis quality, time, tokens, false leads, debugger failures, cleanup, and side effects.

Approach

• Add docs/design/DAP_FAILURE_CLASS_BENCHMARK_PROTOCOL.md.
• Define logs and tests, on-demand DAP, and combined arms with a fixed failure corpus.
• Freeze runtime, model, tool, permission, verifier, and timeout variables.
• Record diagnosis rubric, DAP safety boundaries, overhead, failures, and a decision gate.
• Keep the first step protocol-only; mark DAP not-run when unavailable.

Validation

• git diff --check
• bun run lint
• bun run typecheck

Impact

Documentation only. This is a focused research slice for #168 and contains no DAP runtime or experimental result.

Addresses #168

[tt-a1i]

Thank you for writing down this candidate. We are closing it under the future-capability disposition in #277. The PR remains the preserved research source, but OpenPI will not promote a protocol-only or candidate design into a repository contract before there is an active implementation or experiment with a fixed owner, revision, verifier, and acceptance evidence. This is not a content rejection, and no rebase is requested. Restart the work from the linked source Issue when that evidence frontier is ready.