chore(memory): record typecheck-verification gap and /tmp write flakiness

Verify subagents on a large test-refactor workflow self-reported clean
after lint+test but missed a TS6133 unused-parameter error only
tsc --build caught -- future verify passes on TS changes must run
typecheck too. Also notes that files written under /tmp during this
session intermittently disappeared between tool calls; a real Windows
temp path proved stable instead.

Signed-off-by: Rhuan Barreto <rhuan@barreto.work>

Author: rhuanbarreto

.claude/agent-memory/archgate-developer/MEMORY.md

@@ -24,9 +24,11 @@ Exceptions: minor follow-up tweaks after validation already passed, and non-code
 - **Fire-test a guard in BOTH directions** — that it blocks the bad case AND still permits the legitimate one. A green suite proves only that the gate closes, not that it isn't over-rejecting.
 - **Verify a review agent's claim before acting on it.** They misquote both ADRs and the files they have just read; `grep` the exact quoted string first. A governance finding citing no ADR cannot block on governance grounds — but a demonstrated defect blocks on its own merits.
 - **Content filtering blocks policy/legal boilerplate** — generating a Contributor Covenant or license text can trip API filtering. Ask the user to copy it from the official source.
+- **Files written under `/tmp` by this agent's own Bash/Write calls can vanish between tool calls** — observed for several scratch files with no deleting command run. Write anything that must survive several calls to a real Windows path instead (e.g. `C:/Users/<user>/AppData/Local/Temp/<task-name>/`); Bun/Node on Windows don't resolve Git-Bash-style `/c/Users/...` paths.
 
 ## Topic files
 
+- [Verify agents on TS changes must typecheck](feedback_verify_agents_run_typecheck.md) — `bun test`+lint+format missed a `noUnusedParameters` error a subagent self-reported as clean
 - [Pick the right enforcement layer](feedback_prefer_tests_over_adr_rules.md) — syntax → lint rule; behaviour → test; governance → ADR rule; CLI behaviour → built-in
 - [Answer every review finding on its own thread](feedback_reply_on_review_threads.md) — declines especially; a summary comment does not close the loop
 - [Throw UserError in boundary-wrapped guards](feedback_throw_usererror_in_guards.md) — not `logError` + `exitWith(1)`

.claude/agent-memory/archgate-developer/feedback_verify_agents_run_typecheck.md

@@ -0,0 +1,11 @@
+---
+name: verify-agents-run-typecheck
+description: Verify/review subagents on TypeScript changes must run typecheck, not just lint+test
+metadata:
+  type: feedback
+---
+
+A subagent verifying a TS change must run `bun run typecheck`, not just `bun run lint` + `bun run test` + `bun run format:check`.
+
+- During a 41-file `test.each`/`describe.each` conversion, 7 of 14 fix+verify agent pairs self-reported clean after lint+test but missed `TS6133: 'label' is declared but its value is never read` — a destructured `test.each` row param used only in the title's `%s` substitution, not the callback body. `noUnusedParameters: true` catches it; neither oxlint nor `bun test` does.
+- Still run the full `bun run validate` yourself as the final gate regardless of what subagents reported — self-reported per-file checks miss project-wide issues a full `tsc --build` surfaces.