Update domain-discovery v3.1: Durable Streams feedback

[KyleAMathews]

Summary

Incorporates 9 improvements from the Durable Streams test run feedback (discussion #2):

• Flexible domain count — replace rigid "4-7" target with complexity-driven minimum
• Primary audience identification — group for app devs vs implementors vs operators (prevents mixing protocol internals into feature developer skills)
• Developer task validation pass — merge skills if a developer would commonly need both for one task
• Peer dependency doc reading — new step 7 in reading order; framework integration failure modes were the 6 highest-impact findings in the Durable Streams test but are invisible from the library's own docs
• Wrong/correct code patterns — wrong_pattern and correct_pattern fields in failure mode YAML for direct tree-generator consumption
• source_type field — distinguish documented vs implicit_knowledge failure modes; require Phase 4c to produce 3+ new failure modes not found in Phase 3
• Research-backed interview questions — investigate code before asking maintainer ("I found X — is this intentional?" beats "What happens when X?")
• Pre-1.0 library adaptations — shorter implicit-knowledge phase, longer gap-targeted when usage patterns haven't crystallized
• Composition safety warnings — explicitly probe whether similar-looking APIs from popular libraries transfer correctly

Test plan

• Verify all 9 changes are present in the diff
• Check Phase 3 section numbering is sequential (3a through 3j)
• Verify YAML schema includes new fields (source_type, wrong_pattern, correct_pattern)
• Verify skill_spec.md template reflects new failure mode table columns
• Verify constraints table includes all 9 new checks

🤖 Generated with Claude Code

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/TanStack/playbooks/@tanstack/playbooks@7

commit: e545ff2