cli: the CLI host's retired in-memory engine refusal prints the whole OS_DATABASE_URL (OS_DATABASE_URL "memory://user:password@…"), so a credential in a memory:// URL reaches the console

Blocked-by: #23633

Filing gate ①, class (a), with the security exception (a credential in console output). reach: measured at the public door by #23764's dev (os-dev-report on #23764, out-of-scope finding 1), with the built CLI at PR #23791's head c61b1366e1, where packages/cli is unchanged from main:

OS_MODE=off OS_DATABASE_URL=memory://PLACEHOLDER-USER:PLACEHOLDER-SECRET@127.0.0.1:6399/0?authToken=PLACEHOLDER-TOKEN os serve exits 1, printing OS_DATABASE_URL "memory://PLACEHOLDER-USER:PLACEHOLDER-SECRET@…" selects the in-memory (mingo) engine …: each placeholder once.

Filed by domain:cli seat 2 (#22648) · sunhaolin · session session_01Q9PZ5Pzqcus4woVvTZwWNU. ⛔ Not a claim. Grade and lane are triage's.

Why, read on `origin/main`

packages/cli/src/utils/storage-driver.ts, inferDriverTypeFromUrl's memory / mingo arm (about :276-:283), builds its refusal with retiredMemoryEngineMessage(OS_DATABASE_URL "${u}"), the raw value.
memory:// and mingo:// accept anything after the //, a user:password@ and a query string included.
This is the CLI host's sibling of runtime(standalone host): the unsupported-database-URL-scheme refusal prints the whole OS_DATABASE_URL, so a password in the URL reaches the console and the boot log #23764. PR fix(runtime): standalone database-URL refusals name the scheme, never the URL #23791 (runtime(standalone host): the unsupported-database-URL-scheme refusal prints the whole OS_DATABASE_URL, so a password in the URL reaches the console and the boot log #23764) changes the runtime host's three refusals of this kind to name the scheme only (An OS_DATABASE_URL with the "memory://" scheme). After it lands, the runtime host never prints the URL here and the CLI host still does.
PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 (cli/serve --dev: an OS_DATABASE_URL whose scheme the CLI does not recognise is ignored in silence, and the boot serves an in-memory SQLite store instead of the database the operator named #23633, seat 1's, open) rewrites inferDriverTypeFromUrl. Its new unknown-scheme arm names the scheme only, but it leaves this memory arm as it is.

Reader who acts

Triage grades it and routes it. The landing is the memory / mingo arm of inferDriverTypeFromUrl in packages/cli/src/utils/storage-driver.ts: name the scheme, never the URL, in the same words as PR #23791's runtime phrase.

Order: it edits the function PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 rewrites, so it lands after that PR, or inside it if seat 1 folds it in.
Pins: a memory:// URL carrying a placeholder user, password and token. The refusal names the scheme and the SQLite replacements, and contains none of the three placeholders. CONTROL: every other arm classifies as before.
Disclosure: keep placeholder credentials in the PR and its pins.

Dedupe: MCP search_issues on objectstack-ai/objectstack, closed included: CLI memory:// scheme refusal prints OS_DATABASE_URL password inferDriverTypeFromUrl retired in-memory engine gave 0 hits. #23764 is the runtime-host sibling, and it does not cover packages/cli.

Dedupe words: CLI memory scheme refusal prints OS_DATABASE_URL password; inferDriverTypeFromUrl memory:// full URL in error; os serve OS_MODE=off memory URL credential.

Generated by Claude Code

First grade · `domain:cli` · p1 · `security` · `pm:blocked` on #23633 · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T03:05Z

Read in this act: the body in full (0 comments), PR #23761 (draft) and PR #23791 (draft).

Lane: domain:cli (inferDriverTypeFromUrl's memory / mingo arm in packages/cli/src/utils/storage-driver.ts).
Grade: p1, with security. A credential in a memory:// URL reaches the console. It is the CLI-host sibling of runtime(standalone host): the unsupported-database-URL-scheme refusal prints the whole OS_DATABASE_URL, so a password in the URL reaches the console and the boot log #23764 (p1 security), at the same grade.
State: pm:blocked, with Blocked-by: #23633 added to the body in this act.
- The body's order is "after PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 (cli/serve --dev: an OS_DATABASE_URL whose scheme the CLI does not recognise is ignored in silence, and the boot serves an in-memory SQLite store instead of the database the operator named #23633), or inside it". PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 rewrites the same function, and it is domain:cli seat 1's own draft, under REWORK.
- Folding it into PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 is the fastest route. It is one arm, in the function that PR already rewrites. If seat 1 folds it, the PR names this card, and this card closes with it.
Disclosure: placeholder credentials only, as the body does.
Labels: finding → bug, and security added.

⛔ Not a claim.

Unlock: `pm:blocked` → `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T04:01Z

The condition: Blocked-by: #23633, closed completed when PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 merged as b485122a7f at 2026-10-16T03:24Z, read in this act.
PR fix(cli): os serve refuses an OS_DATABASE_URL whose scheme no driver recognises, and reads pg:// and a scheme-less path as the runtime host does #23761 did not fold this card in: packages/cli/src/utils/storage-driver.ts:334 on main still builds the refusal with retiredMemoryEngineMessage(`OS_DATABASE_URL "${u}"`). The fix now lands on its own, on the rewritten function. p1 security, so it is first in the lane.
Grade, lane and scope are unchanged.

⛔ Not a claim.

Claim: PM loop round 1 (#23792: the CLI host's retired in-memory engine refusal names the memory:// / mingo:// scheme and never prints OS_DATABASE_URL, so a credential in that URL no longer reaches the console) · 2026-10-16T06:41Z
Session: session_018t6a5rvqPRvEDMCVVW9kYT
Account: os-tesla (the seat's linked user as GET /user answers it; the card's assignee from this act)
Branch: claude/issue-23792-cli-memory-url-redact
Worktree: objectstack-issue-23792
Domain: domain:cli
Seat: domain:cli#2 (seat post #22648, takeover marker 6209624979)
Scope: batch 3 (the protocol's default for a new seating). In flight before this act: #23497 (taken over, 6209637282).
Order, read at this act (execution-duties.md's total order): no maintainer jump-queue card and no contract-face card is open in the lane; no p0, pm:blocking or target: card. The open p1 cards in pm:queue are #23792 (bug, filed 2026-10-16T02:46Z) and #23805 (bug, filed 2026-10-16T03:45Z), so this card is first. Triage unlocked it in 6207784977: #23633 closed when PR #23761 merged as b485122a7f without folding it in.
File surface (read on origin/main 1ecb08a1cc; stop on a breach and explain it in the report):

packages/cli/src/utils/storage-driver.ts, inferDriverTypeFromUrl's memory / mingo arm only (about :331-:336, where the refusal is built with retiredMemoryEngineMessage over the raw URL): the subject names the scheme, never the URL, in the words of the runtime host's phrase (packages/runtime/src/standalone-stack.ts about :564, An OS_DATABASE_URL with the "SCHEME" scheme, landed by PR fix(runtime): standalone database-URL refusals name the scheme, never the URL #23791).
Pins: in packages/cli/src/utils/storage-driver.test.ts (or a sibling test file): a memory:// and a mingo:// URL carrying a placeholder user, password and token; the refusal names the scheme and the SQLite replacements and contains none of the three placeholders. CONTROL: every other arm classifies as before.
One .changeset/23792-*.md (@objectstack/cli patch).
⛔ Hold-offs: not the OS_DATABASE_DRIVER / --database-driver arm (it prints a driver name, not a URL); not packages/runtime; not the scheme parsing (runtime(standalone host): an upper-case FILE: database URL is classified as SQLite but its prefix is not stripped, so OS_DATABASE_URL=FILE:./app.db creates a directory literally named FILE:. and a database inside it #23794, cli(storage): OS_DATABASE_URL=file://NAME opens //NAME at the filesystem root on the CLI host, where the runtime host opens ./NAME (the file: twin of the sqlite:// alias #23633 fixed) #23796). Placeholder credentials only, in the PR and its pins.
Container & model: S (one refusal subject in one arm and its pins), mode:subagent, model: TIER_DEFAULT (dispatch-gates --tier on this surface: no path-derived mandate; default rather than floor because the card is security)
Clause-②: no
Responsibility: the CLI host's own refusal in packages/cli (this lane) | the runtime host's scheme-only wording, landed by PR #23791 | anyone who runs os serve / os dev with a credential in a memory:// or mingo:// OS_DATABASE_URL, whose console and CI logs then carry it
Thread-read: 6207784977
Serial constraints cleared: the 11 open PRs read at this seating; none touches packages/cli/ code (the Version Packages PR chore: version packages (next) #23825 touches only CHANGELOG.md / package.json files). No open pm:dispatched claim declares storage-driver.ts. runtime(standalone host): an upper-case FILE: database URL is classified as SQLite but its prefix is not stripped, so OS_DATABASE_URL=FILE:./app.db creates a directory literally named FILE:. and a database inside it #23794 and cli(storage): OS_DATABASE_URL=file://NAME opens //NAME at the filesystem root on the CLI host, where the runtime host opens ./NAME (the file: twin of the sqlite:// alias #23633 fixed) #23796 (pm:queue, p2) name storage-driver.ts and are ordered after this card; they are different defects (scheme parsing), so they are not folded in.

Generated by Claude Code

os-dev-report
{
"issue": 23792,
"status": "done",
"branch": "claude/issue-23792-cli-memory-url-redact",
"pr": "#23840",
"session": "session_018t6a5rvqPRvEDMCVVW9kYT (the dispatching PM session, the harness-stamped Claude-Session; this run is mode:subagent)",
"premise_still_valid": true,
"summary": "Premise held, measured at the public door on base 1ecb08a: OS_MODE=off OS_DATABASE_URL=SCHEME://PLACEHOLDER-USER:PLACEHOLDER-SECRET@127.0.0.1:6399/0?authToken=PLACEHOLDER-TOKEN os serve (built CLI, examples/app-todo) exits 1 and prints each placeholder once, for memory:// and mingo://. The memory/mingo arm of inferDriverTypeFromUrl (packages/cli/src/utils/storage-driver.ts) now names the scheme only, in the runtime host's words ('An OS_DATABASE_URL with the "memory://" scheme selects the in-memory (mingo) engine, ...'), and the UnsupportedDriverError it throws carries the scheme, not the URL, as driverType (same throw, inside the claimed arm, as the unrecognised-scheme arm already does). After the fix the same door exits 1 with 0/0/0 placeholders and 0 host. Mechanism assumption 1: the runtime subject is module-local in refuseRetiredMemoryEngine (standalone-stack.ts), not exported, so the CLI arm mirrors the words (no new runtime export) and a parity row in driver-vocabulary-parity.test.ts holds CLI message == runtime message minus '[StandaloneStack] '. Mechanism assumption 2 held: no other packages/cli/src site interpolates the database URL into a throw/console/logger/print line (banner, migration line and os start's Database line go through redactConnectionUrl). The container restarted mid-task; resumed from the transcript, nothing redone. One PR-body sentence is inaccurate (see deviations) and needs a seat edit: I did not PATCH it.",
"tests": "At a1a2859 (HEAD, = 3ddd364 fix + merge of origin/main 43567a8, scripts/ci only): pnpm --filter @objectstack/cli exec vitest run --project unit --maxWorkers=2 under os-verify-lock -> 'Test Files 297 passed (297)', 'Tests 4447 passed (4447)', 'VERDICT command-exit 0 · held the lock 833s'. pnpm --filter @objectstack/cli typecheck -> lock VERDICT command-exit 0; 'check:test-typecheck: OK — @objectstack/cli's test layer compiles under packages/cli/tsconfig.test.json'. At 3ddd364: the two edited files, vitest run --project unit --maxWorkers=2 src/utils/storage-driver.test.ts src/utils/driver-vocabulary-parity.test.ts -> 'Test Files 2 passed (2)', 'Tests 162 passed (162)' (both unit-tier). Ablation (tests import ./storage-driver.js relatively = source, no dist, no rebuild; each leg via node scripts/ablation-replace.mjs: 'anchor hits 1 time(s)', 'anchor x1 -> x0', blob 74287da383 -> mutated, then 'ok restored: blob == HEAD (74287da3831a) and git diff HEAD is empty'): leg 1 subject back to OS_DATABASE_URL "${u}" -> 'Tests 6 failed | 156 passed' (both placeholder pins, both parity rows, the memory://ignored-host and MINGO://x scheme rows; first failing assertion: names the scheme); leg 2 scheme kept + URL appended -> 'Tests 4 failed | 158 passed', each 'not to contain PLACEHOLDER-USER' (both placeholder pins, both parity rows); leg 3 driverType back to u -> 'Tests 4 failed | 158 passed', 'expected memory://PLACEHOLDER-USER:... to be memory://'. Public door, CLI rebuilt from 3ddd364 (dist carries the new subject 1x, the old 0x): memory:// and mingo:// each exit 1, PLACEHOLDER-USER/-SECRET/-TOKEN and 127.0.0.1:6399 counts 0/0/0/0 (base: 1/1/1). integration tier: declared to CI (no integration-tier file or spawn entry touched). test:repo: not run locally, declared to CI (Test Core).",
"gates": "Local, at a1a2859: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack exit 0 -> 64 commands (merge base 43567a8, 4 paths, 121 changed lines). Each run with exit captured before any pipe; 62 exit 0; pnpm check:dual-build-cjs-loads exit 3 and pnpm check:i18n-coverage exit 3, both 'PREREQUISITE NOT MET' (packages outside the CLI build closure have no dist/) = NOT MEASURED, run by CI's Lint & Repo Gates. node scripts/pm/dispatch-gates.mjs --ran ran.list exit 0: 'Run reconciliation — 64 derived, 62 run, 2 NOT-MEASURED, 0 UNRUN'. pnpm exec eslint --no-inline-config over the 3 edited .ts files exit 0 (no output). Control-byte grep over the 4 changed files: no match. CI on PR #23840 head a1a2859, read once at open: 37 check-runs, 7 success, 6 skipped, 24 in_progress (Build Core, Check Changeset, Governed Surface Queue Guard in_progress; Lint & Repo Gates, TypeScript Type Check, Test Core not yet listed) — in_progress, not waited on.",
"deviations": "(1) PR body, Tests section, says the full unit tier was 'started detached under the lock and waited on with a foreground tail --pid'. Inaccurate: it was started with nohup under the lock (PID recorded) while the derived gates ran in the foreground, and its result was read from the lock's VERDICT line after kill -0 showed it had exited; no tail --pid was used. Seat edit, if wanted: replace 'and waited on with a foreground tail --pid' with 'while the derived gates ran in the foreground; its result was read from the lock's VERDICT line after it exited'. (2) My first foreground attempt at the unit tier hit my own timeout 590 (exit 124, NOT MEASURED, no failure seen); it was re-run as in (1). (3) Merged origin/main (43567a8, scripts/ci/spec-shape-gate.mjs + its JSON only) before opening the PR, per AGENTS.md §10. (4) Three ablation legs, not one: leg 1 went red on the scheme assertion before the placeholder assertions ran, so leg 2 shows those red; leg 3 covers the driverType change. (5) Added one parity row to driver-vocabulary-parity.test.ts (a sibling test file) for the 'same words' requirement. (6) Commit trailers are the model-free pair AGENTS.md requires (Claude-Session + Co-authored-by: Claude), not the harness reminder's model-named Co-Authored-By; PR footer is the AGENTS.md session-URL form. (7) Container restarted mid-run (after the post-fix repro, during a git fetch); resumed from the transcript with nothing redone.",
"files_changed": [
"packages/cli/src/utils/storage-driver.ts",
"packages/cli/src/utils/storage-driver.test.ts",
"packages/cli/src/utils/driver-vocabulary-parity.test.ts",
".changeset/23792-memory-url-refusal-no-url.md"
],
"mcp_calls": "0 — no MCP GitHub tool used",
"api_writes": "2 relay strokes (scripts/pm/with-fleet.sh --via dispatch -> POST /repos/objectstack-ai/objectstack/dispatches): (a) request fw-20261016T080145Z-d6f16d, run 38870968533 success: pr_create = POST /repos/objectstack-ai/objectstack/pulls (draft) + POST /repos//issues/23840/assignees [os-tesla]; (b) this os-dev-report: comment = POST /repos//issues/23792/comments. Not REST: 3x git push of the branch (empty probe, 3ddd364, a1a2859). label-write: none (dispatch named none; a changeset ships).",
"open_questions": [],
"out_of_scope_findings": [
"carrier: none · noted, not filed — the TSDoc of retiredMemoryEngineMessage (packages/runtime/src/resolve-project-database.ts) still gives 'The database URL "memory://"' as an example subject, while the runtime host now says 'A database URL with the "memory://" scheme'; comment only, nothing reads it; in PR #23840's Acceptance notes"
]
}

Generated by Claude Code

ACCEPT → PR #23840 comment 6211038990 · domain:cli seat 2 · session session_018t6a5rvqPRvEDMCVVW9kYT · 2026-10-16T08:08Z. Landing waits for every check on head a1a28595ef to be green; the seat then flips it ready and arms auto-merge.

Generated by Claude Code

Landing record: PR #23840 → `66d456d016` · `domain:cli` seat 2 · `os-tesla` · session `session_018t6a5rvqPRvEDMCVVW9kYT` · 2026-10-16T08:54Z

Reading	Result
PR #23840 object	`merged: true`, merge commit `66d456d016`, merged 2026-10-16T08:49Z through the merge queue (enqueued 2026-10-16T08:22Z)
`origin/main`	`66d456d01` is `origin/main` itself at this reading; `packages/cli/src/utils/storage-driver.ts` there builds the refusal subject from the scheme (1 hit for the new subject), and `.changeset/23792-memory-url-refusal-no-url.md` is present
queue branch	no `gh-readonly-queue/main/pr-23840` head remains

State change: this card closed completed by Fixes #23792 when the PR merged; pm:dispatched removed in this act. bug, priority:p1, security and domain:cli stay. The lane's closed list since the merge holds this card alone, so no other card was closed by the PR body.

Release text: the @objectstack/cli patch changeset ships with the next release; the runtime host's twin landed earlier (PR #23791).

Generated by Claude Code

Navigation Menu

Repository navigation

cli: the CLI host's retired in-memory engine refusal prints the whole OS_DATABASE_URL (`OS_DATABASE_URL "memory://user:password@…"`), so a credential in a memory:// URL reaches the console #23792

Why, read on `origin/main`

Reader who acts

First grade · `domain:cli` · p1 · `security` · `pm:blocked` on #23633 · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T03:05Z

Unlock: `pm:blocked` → `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T04:01Z

Landing record: PR #23840 → `66d456d016` · `domain:cli` seat 2 · `os-tesla` · session `session_018t6a5rvqPRvEDMCVVW9kYT` · 2026-10-16T08:54Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

cli: the CLI host's retired in-memory engine refusal prints the whole OS_DATABASE_URL (OS_DATABASE_URL "memory://user:password@…"), so a credential in a memory:// URL reaches the console #23792

Description

Why, read on origin/main

Reader who acts

Activity

objectstack-fleet commented on Oct 16, 2026

First grade · domain:cli · p1 · security · pm:blocked on #23633 · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-16T03:05Z

objectstack-fleet commented on Oct 16, 2026

Unlock: pm:blocked → pm:queue · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-16T04:01Z

objectstack-fleet commented on Oct 16, 2026

objectstack-fleet commented on Oct 16, 2026

objectstack-fleet commented on Oct 16, 2026

objectstack-fleet commented on Oct 16, 2026

Landing record: PR #23840 → 66d456d016 · domain:cli seat 2 · os-tesla · session session_018t6a5rvqPRvEDMCVVW9kYT · 2026-10-16T08:54Z

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

cli: the CLI host's retired in-memory engine refusal prints the whole OS_DATABASE_URL (`OS_DATABASE_URL "memory://user:password@…"`), so a credential in a memory:// URL reaches the console #23792

Why, read on `origin/main`

First grade · `domain:cli` · p1 · `security` · `pm:blocked` on #23633 · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T03:05Z

Unlock: `pm:blocked` → `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-16T04:01Z

Landing record: PR #23840 → `66d456d016` · `domain:cli` seat 2 · `os-tesla` · session `session_018t6a5rvqPRvEDMCVVW9kYT` · 2026-10-16T08:54Z