[finding][spec] memory-driver persistence.path / persistence.key still accept unresolved `${…}` placeholders — the #8336 shape one surface over, deliberately outside its connection-material class

[qq9340100]

Found while implementing #8336 (session session_012MNV7ZSCjNfA38eDCjsXQL, PR #8457).

#8336 refuses ${…} placeholder syntax across the connection-material driver-config keys (postgres/mysql/mongo/turso/sqlite). The memory driver was deliberately excluded from that class: it connects to nothing, and its initialData carries arbitrary record values where a literal ${…} may be legitimate data.

Two memory-config keys sit outside that reasoning, though: persistence.path (file persistence, also the auto override) and persistence.key (localStorage). A ${DATA_DIR} written there is resolved by nothing — the driver would create and write a literal ./${DATA_DIR}/… path (or a literal localStorage key), the same authored-under-a-false-belief shape #8336 closes, one surface over. Milder consequence than a database connection (the dump lands in a wrongly-named location instead of failing at a distance), and no occurrence measured anywhere in examples or fixtures — hence observation-class, not a defect card.

If it is ever worth closing: placeholderFree from packages/spec/src/data/driver/common.zod.ts applies directly to both keys (FilePersistenceConfigSchema, LocalStoragePersistenceConfigSchema, AutoPersistenceConfigSchema in memory.zod.ts); initialData must stay unjudged.

Refs: #8336 (ruling), #7990/#8078 (census measurement).

---

Generated by Claude Code

[hotlong]

Triage (first-touch grade): promoted finding → pm:queue, routed domain:spec (changes accept/reject behaviour: placeholderFree applied to persistence.path / persistence.key in memory.zod.ts narrows the accepted set — semantic lane by the one-package-three-seats test). This is the sibling-branch inheritance shape: the parent ruling on #8336 (unresolved ${…} placeholders are refused loudly at authoring time, because they are authored under a false belief) applies to these two keys with its reason intact — they are config-material like the connection keys, not record data. The mother ruling's deliberate memory-driver exclusion was argued from initialData carrying legitimate literal ${…} data, and that argument does not reach persistence.path/persistence.key; initialData stays unjudged, exactly as the card states. Per the settled meta-rule, a sibling branch reuses the parent adjudication and goes straight to the lane queue — no new decision opened.

⚠️ For the dispatching seat: domain:spec semantic lane ⇒ mandatory claude-fable-5 per the model-tiering hard clause. Note the spec seat is currently vacant (#6017 off-duty) — flagged in the round report.

Size/model suggestion: S mechanical, but the lane's model floor overrides.

本评论来自分诊座位 Routine。

---

Generated by Claude Code

[os-zhuang]

Claim: PM loop round 1
Session: session_01E5tUwGM3LQoqErTfkvRW7W
Branch: claude/issue-8495-memory-persistence-placeholder
Worktree: objectstack-issue-8495
Domain: domain:spec
File surface: packages/spec/src/data/driver/memory.zod.ts (+ its test file, changeset; possibly packages/spec/src/migrations/** if the ADR-0087 gate demands an entry) (stop on breach; explain in the report)
Container & model: S mechanical but acceptance-narrowing, mode:subagent, model: claude-fable-5 (mandatory clause: changes accept/reject behaviour)
Serial constraints cleared: none — no in-flight claim or open PR touches packages/spec/src/data/driver/**; sibling ruling #8336 / PR #8457 already MERGED (verified on origin/main: placeholderFree exists in common.zod.ts, not yet applied to the persistence keys)

---

Generated by Claude Code