fix: Drain GitHub issue hasna/sessions#17 for todos task a1b9689e. Issue

[andrei-hasna]

Objective

Drain GitHub issue #17 for todos task a1b9689e. Issue URL: #17. Read the supplied context file. Prior run run_06c012c7d273 passed install/typecheck/build/test but stopped before commit because the staged secrets scan found three secret-like findings in staged blobs; do not read or print those values. Fix the candidate so staged secrets scan passes, using recognizable non-live placeholders plus hasna:allow-secret comments only for deliberate test fixtures, or remove any real secret material. Then open a PR only after required gates support it. Not drained until PR merges and issue #17 closes.

Verification

• policy source: base 5454297 (immutable commit — agent-proof)
• containment: env — allowlist env, non-login shell, run-scoped HOME (registry auth seeded for install)
• install: pass
• typecheck: pass
• build: pass
• test: pass
• doctor (ci): ok — 11 checks passed (1 advisory)

---

Run run_cd7f0a4bbd92 · backend codewith · task inline
🏭 Generated by @hasnaxyz/factory

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[andrei-hasna]

[REVIEW] GO — #52 @ 9d821aa — lens: correctness+security+gates, reviewer unresolved-account001 (1 of 1)

Commands run at the reviewed head:

• bun install — exit 0; setup only, 117 packages installed (not reported as a test gate).
• bun run typecheck — exit 0; 0 TypeScript diagnostics.
• bun run test — exit 0; 374 pass, 0 fail, 1,771 assertions across 43 files.

Read and traced:

• git log --oneline origin/main..HEAD, git diff origin/main...HEAD --stat, and the full origin/main...HEAD diff for all 8 changed files, using fetched base 545429799435cae2e42428c8d6b3fcfa70ba6c9d.
• Full surrounding ingestion parser, state, database, and watcher source, including the Codewith/Codex parser wrappers and the complete changed ingestion test.
• PR metadata and issue Preserve Codewith titles parentage and provenance during ingestion #17 acceptance criteria: indexed titles, structured parent/subagent provenance, retained timestamps/model/provider/project/machine/source path, and exclusion of auth-profile/credential metadata.

Blocking P0/P1 findings: none. The session-index parser copies only id and thread_name; structured source metadata is allowlisted; auth-profile and credential fields are not ingested or emitted; auxiliary index changes invalidate stored ingestion state; before/after signatures defer racing writes rather than persisting mixed snapshots.

Non-blocking follow-up (P2 test hardening): the new assertions use toMatchObject, so they would not catch future extra auth/credential keys in stored metadata. Current implementation is safe, but exact metadata equality or explicit absence assertions would make that invariant regression-proof.