The shipped contacts.csv import template fails 6 of its 50 rows: `Department: "HR" is not a known option` (the option is `hr`, labelled "Human Resources")

[objectstack-fleet]

Filing gate: ① product defect with reach measured. Class (a). reach: public door. POST /api/v1/data/crm_contact/import (mappingName: crm_contact_import) with the shipped assets/import-templates/contacts.csv, on @objectstack/* 17.6.0: the dry run gives 50 total / 44 ok / 6 errors, and the commit gives 44 created / 6 errors. All six read Department: "HR" is not a known option. They are rows 7, 15, 23, 31, 39 and 47.

Who acts on it: the repo:hotcrm seat, which dispatches a dev. Found by the seat's dev while converting #1836 (session_01ER8ntXZhYebyQ66aXWdjfT). It was already noted on #1836 in R64 (5833812920) and never filed. ⛔ Not a claim; the first-touch grade is owed.

Measured

• The count is the same on main 4807ed9e and on the crm_contact models a postal address as five flat text fields while crm_account and crm_lead use Field.address() — and AGENTS.md's own guidance names Field.address() for exactly this #1836 conversion branch (PR feat(contact): one structured mailing_address replaces five flat mailing_* fields (#1836) #1997), so it predates the conversion and is independent of it.
• crm_contact.department declares the option value hr, label "Human Resources". The template's cell is HR, which matches neither.
• The template is what a new customer imports first. A 12% row-failure rate on the vendor's own sample file is the first impression of the import path.

Fix space (for the dev, not decided here)

• Option A: map HR → hr in contact_import.mapping.ts, the same synonym mechanism LEAD_SOURCE_SYNONYMS uses. This keeps the customer-facing template unchanged, which is the contract crm_contact models a postal address as five flat text fields while crm_account and crm_lead use Field.address() — and AGENTS.md's own guidance names Field.address() for exactly this #1836's ruling protected.
• Option B: change the template's cells to Human Resources.
• Either way, pin it with a test: the shipped template imports with 0 errors.

Duplicate check

All hotcrm issues, open and closed: 748 issues over 19 REST pages, read to the short page. Title and body were grepped for (contacts.csv|import template)…(HR|department)|Department: "HR"|"HR" is not a known: 0 hits. Positive control: contacts.csv hit 3 (#1836, #1827, #1199).

Dedupe words: contacts.csv HR department invalid option import template 6 rows

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round R72
Session: session_01ER8ntXZhYebyQ66aXWdjfT
Account: hotlong (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-1998-contacts-csv-hr-department
Worktree: hotcrm-issue-1998
Domain: repo:hotcrm (single-lane repo, no domain:* taxonomy)
Seat: repo:hotcrm#1
File surface: src/sales/mappings/contact_import.mapping.ts and src/sales/mappings/_shared.ts (a department synonym map beside LEAD_SOURCE_SYNONYMS); one test that imports the shipped template; one .changeset/1998-*.md. ⛔ assets/import-templates/contacts.csv does not change: it is the customer-facing contract #1836's ruling protected (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: default (dispatch-gates --tier --repo objectstack-ai/hotcrm answers "no path-derived mandate … the tier stays the PM's per-card judgment call")
Clause-②: no. This is an app import mapping; it touches no published schema and no accept set.
Thread-read: none
Serial constraints cleared: PR #1997 (the contact mapping's address targets) merged at f44ab642. #1999 touches only scripts/backfill-*.ts, which is disjoint. No open PR exists.

Seat's first-touch grade: p2 bug, dispatched directly (pm:dispatched). The reach is measured at a public door. The fix is the mapping synonym (option A on the card), which keeps the template unchanged.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report
{
"issue": 1998,
"status": "done",
"branch": "claude/issue-1998-contacts-csv-hr-department",
"pr": "#2001",
"session": "session_01ER8ntXZhYebyQ66aXWdjfT (subagent = the parent seat's id)",
"premise_still_valid": true,
"summary": "The contact import mapping now sends Department through transform 'map' with a new DEPARTMENT_SYNONYMS = { 'HR': 'hr' } in src/sales/mappings/_shared.ts, which is the LEAD_SOURCE_SYNONYMS mechanism. assets/import-templates/contacts.csv is byte-unchanged (git diff f44ab64..HEAD -- assets/ is empty). The root cause, read in @objectstack/core 17.6.0: matchOption compares an option VALUE exactly and only the LABEL case-insensitively, so "HR" matched neither "hr" nor "Human Resources". I corrected the _shared.ts header, which claimed that values match case-insensitively; that file is in the claim's file surface. The template's Department column holds 8 distinct values. 7 of them are option labels (Executive, Sales, Marketing, Engineering, Support, Finance, Operations) and land unaided, so the map has exactly one entry. A new it.each in test/import-mappings.test.ts runs every row of all three shipped templates through the import door's own applyMappingToRows + coerceRow (@objectstack/core) and expects no cell error. accounts.csv and leads.csv were already clean. PR #2001 is a draft, assigned to hotlong. It also carries a ci/cd label that the Label Pull Request workflow applied through the labeler.yml test/** glob; I did not write it.",
"tests": "pnpm verify under the lock (OS_VERIFY_LOCK_SLOT=hotcrm-issue-1998 bash /home/user/objectstack/scripts/pm/os-verify-lock.sh -c 'pnpm verify') at HEAD f10cae6 (git rev-parse --short HEAD, same run): "os-verify-lock: VERDICT command-exit 0 · held the lock 183s (3m03s) · waited 0s". Inside it: "✓ Validation passed", typecheck clean, lint "1 warning(s), 18 suggestion(s)" (the warning is the existing sales_home_page page:card; none is on a mapping), "✓ i18n lint gate: 0 i18n/missing-* issues", "✓ source hygiene clean", "✓ source token ratchet clean", build ok, "Test Files 176 passed (176)", "Tests 3790 passed | 1 skipped (3791)". Single file, verbose: test/import-mappings.test.ts "Tests 42 passed (42)", including the 3 new tests "every cell of 'accounts.csv' / 'contacts.csv' / 'leads.csv' coerces through ... with no error". ABLATION, run from committed b66111a with /home/user/objectstack/scripts/ablation-replace.mjs in wrap mode (--anchor " 'HR': 'hr',\n" --delete, around the vitest run under the lock): "ok mutation landed: anchor 1 → 0, blob b8616d6d2395 → 8ec71c8b8058"; then "Tests 1 failed | 41 passed (42)". The failing test is "every cell of 'contacts.csv' coerces through 'crm_contact_import' with no error", and it lists rows 7, 15, 23, 31, 39 and 47 as "department invalid_option: Department: \"HR\" is not a known option". Restore: "ok restored: blob == HEAD (b8616d6d2395) and git diff HEAD is empty", and git status --short was empty. The direction was red, as predicted. There is no build/dist leg because vitest resolves the mapping from src through objectstack.config.ts. ENDPOINT, measured with objectstack start -p 4818 on a fresh worktree DB, after a local sign-up and importing accounts.csv first (50/50 created). BEFORE (f44ab64): POST /api/v1/data/crm_contact/import with mappingName crm_contact_import gave total 50, ok 44, errors 6 on the dry run (default), the same on the dry run with runAutomations:false, and created 44, errors 6 on the commit. All six were "Department: \"HR\" is not a known option" on rows 7, 15, 23, 31, 39 and 47. AFTER (b66111a): both dry runs gave total 50, ok 50, errors 0, and the commit gave created 50, errors 0. Reading back the six HR contacts by email gave department "hr"; control row 1 (ana.haddad@...) read back "executive". The served mapping (GET /api/v1/meta/mapping/crm_contact_import) had Department {"transform":"map","params":{"valueMap":{"HR":"hr"}}}, equal to dist/objectstack.json from the same build; before, it was transform "none". Both servers were stopped by recorded PID, and port 4818 is free. CI at report time on f10cae6: Check Changeset, Quality Checks, Label Pull Request, link-check and Vercel Preview Comments succeeded. Build and Test (22.x), Playwright and Analyze Code were in_progress.",
"token_ratchet": "src/sales, run before and after with node scripts/check-source-token-ratchet.mjs. BEFORE: "other authored metadata 45 1,296 61,985 15,496 —" and "authored total 117 10,409 399,528 99,882 107,000". AFTER: "other authored metadata 45 1,304 62,174 15,544 —" and "authored total 117 10,417 399,717 99,929 107,000". Business semantics 56,418 / 59,000 and interaction layer 27,968 / 31,000 are unchanged. No ceiling was touched, and the ratchet's advisory to re-anchor the interaction ceiling to ~30,000 was not acted on.",
"files_changed": [
"src/sales/mappings/_shared.ts",
"src/sales/mappings/contact_import.mapping.ts",
"test/import-mappings.test.ts",
".changeset/1998-contacts-template-hr-department.md"
],
"deviations": [
"Hypothesis 2 read literally ("a synonym for EVERY template value that is not already an option value") would add 7 entries: Executive, Sales and the others are not option VALUES, but each one equals an option LABEL, and the platform label match already lands it. Those entries would be dead, against the ban on synonyms the import does not need, so the map holds only HR. I read the instruction as "every template value that matches no option".",
"The _shared.ts header comment was corrected; the template file was not touched. The comment said that option values match case-insensitively, which is the belief that hid this defect. Left uncorrected, it would let the next agent delete the HR entry as redundant.",
"Measurement auth: objectstack start seeds no admin@objectos.ai (sign-in answered 401), and a request without an Origin header answers 403 MISSING_OR_NULL_ORIGIN. So the script signed up a local user in the disposable worktree DB and sent Origin. The import ran as that user."
],
"mcp_calls": "0 — none",
"api_writes": "3 — each went through the fleet-write relay as one POST /repos/objectstack-ai/objectstack/dispatches: (1) pr_create → POST /repos/objectstack-ai/hotcrm/pulls (#2001, draft forced, 6027 bytes read back identical); (2) label-write.mjs --issue 2001 --assign hotlong → POST /repos//issues/2001/assignees (read-back MATCHES: labels ci/cd, assignees hotlong); (3) this os-dev-report → POST /repos//issues/1998/comments. Plus 4 git pushes (the empty-branch probe and 3 commits), which are not REST.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: the customer import guides over-claim picklist matching. content/docs/guides/importing-your-data.mdx says "accept HotCRM's own labels or codes, case-insensitively", and content/docs/guides/import-and-export.mdx says "matched against the field's option labels and values, case-insensitively"; the zh-Hans and zh-Hant copies of each say the same. In @objectstack/core 17.6.0, matchOption compares a code exactly, so an upper-cased code fails its row. I read this from source and measured it at the door only for HR, which this PR now maps. · carrier: none (承接者:无) · noted in PR 2001 Acceptance notes, not filed · dedupe words: import guide picklist case-insensitive code value exact match matchOption",
"carrier: the JSDoc in src/sales/mappings/_shared.ts points INDUSTRY_SYNONYMS at src/objects/_picklists.ts, but the file is src/sales/objects/_picklists.ts (a stale path in a comment) · carrier: none (承接者:无) · noted in PR 2001 Acceptance notes, not filed"
]
}

---

Generated by Claude Code

[objectstack-fleet]

ACCEPT: PR #2001 at f10cae68. The shipped contacts.csv imports all 50 rows. repo:hotcrm seat, session_01ER8ntXZhYebyQ66aXWdjfT, 2026-10-03T09:50Z. Report: the dev's os-dev-report 5967929249.

Probes taken by the seat itself:

• Ancestry and checks: main f44ab642 is an ancestor of f10cae68. CI 9/9 success. Governed check: 0 of 4 paths, NOT governed; 105 lines.
• Template untouched: git diff origin/main…f10cae68 -- assets/ is empty, so crm_contact models a postal address as five flat text fields while crm_account and crm_lead use Field.address() — and AGENTS.md's own guidance names Field.address() for exactly this #1836's template contract holds.
• Read line by line: DEPARTMENT_SYNONYMS = { 'HR': 'hr' } beside LEAD_SOURCE_SYNONYMS. The Department column goes through transform: 'map'. The _shared.ts header now states the measured matching rule: the value matches exactly, the label case-insensitively. The old text claimed the value matched case-insensitively too, and that belief hid this defect. The new test runs every row of all three shipped templates through the import's own applyMappingToRows + coerceRow.

The dev's measurements, accepted:

• Endpoint on 17.6.0: before the fix, the dry run and the commit each give 44 ok / 6 errors (rows 7, 15, 23, 31, 39, 47). After it, both give 50 / 0. The six read back department: 'hr'.
• Ablation: deleting the entry turns exactly the contacts-template test red, listing the six rows; the restore is proven.
• Gates: pnpm verify VERDICT command-exit 0 (176 files). Tokens: src/sales authored total +47, no ceiling moved.
• The map holds one entry, not eight. The other seven template values equal option labels and land unaided, so adding them would only add dead synonyms.

Noted for a follow-up, not blocking:

• The customer import guides (content/docs/guides/importing-your-data* and import-and-export*, in 3 locales) still say codes match case-insensitively. 17.6.0's matchOption compares a code exactly.
• One JSDoc path in _shared.ts is stale.
• Both are in the PR's Acceptance notes.

Landing:

• Authority: 「授权你执行pr合并」 (2026-10-03, about 04:52Z), on a p2 bug card the seat filed with reach measured.
• Path: ready → auto-merge → queue.

---

Generated by Claude Code