fix(client create): strip control chars from --name before slugifying

[LukasWodka]

Summary

CLI-side defense-in-depth for the name-garble bug (customer-reported 2026-07-20) — the companion to the installer fix in tracebloc/client#362.

Typing arrow keys at the installer's name prompt injected raw ESC[D/ESC[A bytes into the client name. slug.Slugify keeps ESC (0x1B < 128) through its ASCII pass and then collapses each ESC[<x> run into a -, minting the exact garbage namespace the customer saw:

"se-\e[D\e[D\e[A\e[A" → slug → "se-d-d-a-a"

The installer now strips at the source (client#362), but a --name / $TRACEBLOC_CLIENT_NAME can also reach client create directly, so this hardens the CLI boundary too.

Changes

• internal/cli/sanitize.go — new sanitizeClientName(): strips ANSI CSI sequences (arrow keys, cursor moves, bracketed-paste wrappers), the literal [200~/[201~ paste markers, and any remaining C0/DEL control chars. UTF-8 (≥ 0x80) preserved. Mirrors the installer's _strip_paste_garbage.
• internal/cli/client.go — runClientCreate cleans both name and location right after resolving them, before the auto-name check and before slug.Derive. A name that is only control chars cleans to "" and falls through to auto-naming (the same graceful path as an omitted --name). Both cleanups log before/after to the install trace; a non-empty change also prints a one-line notice.

Why not fix it in slug.Slugify?

slug must stay a faithful mirror of the backend's slug.py — the backend validates exactly what it produces (RFC-0001). Stripping there would diverge the two. Input hygiene belongs at the CLI ingestion boundary, not in the shared slug rule.

Test plan

• internal/cli/sanitize_test.go — 15 cases: arrow keys, Ctrl+arrow (CSI params), Delete (ESC[3~), paste wrappers, post-corruption literal markers, bare C0/DEL, lone ESC, and UTF-8 / legitimate-bracket preservation.
• End-to-end assertion: the raw input slugifies to the documented se-d-d-a-a garble and the sanitized input slugifies clean to se — so a future slug change that alters either is caught here.
• go build ./..., go vet, gofmt -s, goimports -local, staticcheck (all,-ST1005), deadcode, and file-budget all clean.

🤖 Generated with Claude Code

---

Note

Low Risk
Localized input sanitization on client create flags/env with broad test coverage; no auth, backend contract, or slug parity changes.

Overview
Adds CLI-side defense-in-depth so arrow keys and other terminal junk in --name / $TRACEBLOC_CLIENT_NAME (and --location) no longer produce garbage Kubernetes namespaces like se-d-d-a-a when slug.Slugify runs.

Introduces sanitizeClientName to strip ANSI CSI sequences, bracketed-paste markers, and C0/DEL control bytes while keeping UTF-8. runClientCreate applies it to name and location before auto-naming and slug derivation; control-only names become empty and follow the same auto-name path as a missing --name. Cleaned values are written back into opts immediately so failure-path resumeCommand does not echo raw escapes.

Tests cover sanitization cases, slug end-to-end, and a create failure where garbled input must not reappear in output or the resume line.

^{Reviewed by Cursor Bugbot for commit 1e03eb5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

@BugBot run

[LukasWodka]

👋 Heads-up — Code review queue is at 31 / 30

Above the WIP limit. The team convention is to review existing PRs before opening new work.

Open PRs currently in Code review (oldest first):

• averaging-service#204 — chore(release): promote staging → main (week of 2026-07-20) · author: @aptracebloc · no reviewer assigned
• backend#1131 — perf(psr): offload pause/stop edge notify to Celery (#980) · author: @aptracebloc · reviewer: @saqlainsyed007
• backend#1134 — perf(landing): bulk submission counts + team-member fold (#978 pt5 steps 2-3) · author: @aptracebloc · reviewer: @saqlainsyed007
• backend#1139 — refactor: dedup runner twins, edge-role permissions, LOGGING builder (#961) · author: @saqlainsyed007 · no reviewer assigned
• backend#1147 — chore(release): promote staging → master (week of 2026-07-20) · author: @aptracebloc · reviewer: @saadqbal
• backend#1148 — fix(sb): always close the failed client in reply_to_edge (Bugbot #1147) · author: @aptracebloc · reviewer: @saqlainsyed007
• backend#1149 — chore(models): remove dead Company model (superseded by Account) · author: @saqlainsyed007 · reviewer: @saadqbal
• cli#354 — fix(doctor): stop the cluster-down cascade of misleading remedies · author: @LukasWodka · reviewer: @saadqbal
• cli#356 — fix(doctor): requests-proxy check — result-egress framing + honest ✔ (cli#351 Part 1) · author: @LukasWodka · reviewer: @saadqbal
• cli#358 — fix(home): realistic heartbeat timeout so "· Online" stops flickering to "couldn't confirm" · author: @LukasWodka · reviewer: @saadqbal

Pull from review before opening new work. (This is a nudge from the kanban WIP check, not a block.)

[LukasWodka]

@BugBot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 1e03eb5. Configure here.}

[aptracebloc]

Verified — approving. Solid defense-in-depth input hardening:

• Sanitizer order is correct: whole CSI sequences (incl. their ESC) stripped first, then the literal [200~/[201~ paste markers, then bare C0/DEL last — stripping ESC first would have broken the CSI match and left [D. UTF-8 is preserved (strings.Map operates on runes, so any rune ≥ 0x80 survives). The regex covers the real threat: arrow keys, Ctrl+arrow params, Delete (ESC[3~), and paste wrappers. A control-only name cleans to "" → graceful auto-naming, same as an omitted --name.
• Failure-path echo closed: writing the cleaned values back into optsimmediately means an early return's resumeCommand(opts) hint can't reprint the raw escapes (the resolved "stale opts" case). Nicely subtle.
• Right layer: sanitizing at the CLI boundary rather than in slug.Slugify keeps slug a faithful mirror of the backend's slug.py.
• Strong tests: 15 sanitizer cases plus the e2e assertion that pins both the garble repro (se-d-d-a-a) and the clean result (se).

Two non-blocking micro-gaps (outside the threat model — no change needed):

1. An incomplete CSI (ESC[ with no final byte) isn't matched, so after the ESC is stripped a stray [ remains.
2. A private-mode CSI (ESC[?25h — the ? = 0x3F isn't in the [0-9;] param class) likewise falls through, leaving [?25h after the ESC-strip.

Neither arises from real terminal arrow-key/paste input, both leave only printable chars slug handles (not garbage namespaces), and this is defense-in-depth behind the installer fix (client#362) — so nothing to change here. Flagging only for the record: if you ever want to be exhaustive, broadening the param class (e.g. [0-9;?<>=]*) and/or dropping a trailing lone ESC[ would close them.

LGTM.

— drafted with Claude (Opus 4.8), sent by @aptracebloc

[saadqbal]

/fr-pass