Demo seed: no `ats_offer` row belongs to a hired application — all 14 offers sit on `offer`-stage applications, so "days to offer for hired applications" is unmeasurable on the seed

[claude]

Found while building card 13 (#8). Read from dist/objectstack.json at 8d5dcc2 (the seed is literal rows, so this is a property of the source, not of a boot):

• ats_application: 200 rows, stage counts applied 88 · screening 46 · interview 28 · offer 14 · hired 9 · rejected 15 · withdrawn 0.
• ats_offer: 14 rows. Joining each offer's application back to its application: 14 of 14 point at an offer-stage application, 0 point at a hired one. The skeleton's own invariant says so (src/data/shared/pipeline.ts: "offers: 14 · one per offer-stage application").
• ats_offer.status: sent 5 · pending_approval 3 · approved 3 · draft 2 · declined 1 — no accepted.

Why it matters: the state machine only reaches hired from offer (DESIGN.md §02), and F3 (offer_approval) is what produces the offer row — so a hired application with no offer is a history the app itself cannot produce. Concretely, the employer dashboard's fourth tile in card 13 ("median days from applied_at to the offer's created_at for hired applications") has no rows to read on the demo seed, independently of how it is computed. The 9 hired applications also carry no accepted offer for the offer views to show.

What a fix looks like (not done in #8 — src/data/ is outside that card): give each of the 9 hired applications one ats_offer row with status: 'accepted' (and a start_date), keeping the total offers coherent with DESIGN.md §06 (14 → 23, or re-balance), and add the invariant to the skeleton header. Whether ats_offer needs a stored accepted_at/created_at for the duration measure is the separate decision the #8 report raises.

Reproduce: python3 -c "import json,collections; d=json.load(open('dist/objectstack.json')); s={x['object']:x['records'] for x in d['data']}; st={a['display_name']:a['stage'] for a in s['ats_application']}; print(collections.Counter(st[o['application']] for o in s['ats_offer']))" → Counter({'offer': 14}).

---

Generated by Claude Code

[os-sam]

Claim: session_01PbJ5Cy9KDAzeQHo8bsMadG · branch claude/issue-53-offers-on-hired

main is 0ad34d7 and the repository now has CI — this is the first card whose PR will be gated by a machine rather than by a pasted terminal tail. pnpm lint is also --i18n-strict now, so anything with a new label needs its zh-CN string or the gate fails.

#61 is folded in. It is the same subject — AGENTS.md and a src/data/index.ts comment still say 801 seed rows while the loader, README and boot banner say 809 — and this card moves that number again, so fixing it anywhere else would just make it wrong twice.

This card unblocks #55 (time-to-offer), which cannot be built until a hired application has an offer to measure against.

The ripples the card text does not mention

Changing the seed moves numbers that are now written down in several places, and after #9 they are user-facing:

• README states the seed size in three places and the boot banner prints it.
• AGENTS.md states it too (that is AGENTS.md still says the demo seed is 801 rows; the loader, README and boot banner say 809 #61).
• DESIGN.md §06 carries the per-object seed table.
• The release screenshots (docs/screenshots/) show funnel and dashboard numbers. Stage counts do not change — this card adds offers, not applications — so 88 / 46 / 28 / 14 / 9 stands and the shots stay valid. ⛔ Verify that rather than assuming it; if any screenshot number does move, say so and reshoot it.

One thing to decide, not to guess

The card offers "14 → 23, or re-balance". Either is fine, but say which and why, and keep DESIGN.md §06 and the README consistent with whatever lands. The invariant the skeleton header states (offers: 14 · one per offer-stage application) has to be rewritten to whatever becomes true — an invariant comment that no longer holds is the same defect class this project has spent the day removing.

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 53,
  "folded_in": 61,
  "status": "done",
  "branch": "claude/issue-53-offers-on-hired",
  "pr": "https://github.com/objectstack-ai/ats/pull/64",
  "premise_still_valid": true,
  "summary": "Premise re-measured on origin/main 0ad34d7 and held: 14 ats_offer rows, all on offer-stage applications, 0 of 9 hired applications with an offer, 809 seed rows. Added one `accepted` offer per hired application in the shared skeleton (pipeline.ts, 9 rows, table kept sorted by application index) so both locale packs render it through scopeToDemo(); buildOffers now stamps approved_by on accepted rows and takes a signed day offset because an offer accepted 33 days ago has a start date and expiry already in the past. Rewrote the skeleton header invariant to what is now true. #61 folded in: AGENTS.md 801 -> 818 (twice), src/data/index.ts 794 -> 811; README 809 -> 818 (x3) and its isolation sentence (Quillstone 3 offers, Harborline 5/31/4/3/2), DESIGN.md §06 ats_offer 14 -> 23, seed-file comments, CONTRIBUTING example, and a paragraph in docs/screenshots/README.md. No file on the ⛔ list touched; no new label/option/field (i18n-strict had nothing new).",
  "sizing_decision": "14 -> 23. Every offer-stage application needs its offer for the same state-machine reason the hired ones do (F3 wrote it), so the 14 cannot be redistributed; a re-balance inside 14 would move applications between stages and therefore the 88/46/28/14/9 funnel the release screenshots show. Adding 9 keeps every application count unchanged — measured in the artifact and at runtime.",
  "files_changed": [
    "src/data/shared/pipeline.ts",
    "src/data/shared/build.ts",
    "src/data/demo-en/offer.seed.ts",
    "src/data/demo-zh/offer.seed.ts",
    "src/data/index.ts",
    "AGENTS.md",
    "README.md",
    "DESIGN.md (§06 row only)",
    "CONTRIBUTING.md",
    "docs/screenshots/README.md"
  ],
  "gates": {
    "local_on_54a442c": "pnpm validate exit 0 · pnpm lint exit 0 (3 pre-existing approval-approvers-may-resolve-empty suggestions, unchanged) · pnpm typecheck exit 0; each via os-verify-lock.sh, verdict lines `VERDICT command-exit 0`, exit codes captured before any pipe",
    "ci": "in_progress at report time — https://github.com/objectstack-ai/ats/actions/runs/34149866719 (pull_request event on 54a442c); CI convergence not waited for, per contract"
  },
  "tests": "No unit tests in this repo; verification = the three gates (above) + artifact join + runtime read-back (below). First gate pass ran unlocked (the lock script on this box is mode 644, so an -x check skipped it); the final pass on the committed HEAD ran under `bash os-verify-lock.sh -c`.",
  "artifact_join": {
    "before_0ad34d7": "Counter({'offer': 14}) · hired with offer 0 of 9 · offer status sent 5 / pending_approval 3 / approved 3 / draft 2 / declined 1 · 809 rows / 17 datasets",
    "after_54a442c": "Counter({'offer': 14, 'hired': 9}) · hired with offer 9 of 9 · accepted 9 + the 14 unchanged · every accepted offer on a hired application: true · every non-accepted on an offer-stage application: true · 818 rows / 17 datasets",
    "stage_counts": "applied 88 · screening 46 · interview 28 · offer 14 · hired 9 · rejected 15 — identical before and after",
    "offers_per_employer": "Quillstone 2 -> 3 · Harborline 2 -> 4 (also Cedarbrook 4, Brightmarket 3, Summitridge 3, Swiftroute 3, Lumenvale 1, Bluewater 1, Ironbridge 1)"
  },
  "runtime": {
    "boot": "memory driver, port 4403, counted after the seeder line. before: `loading 17 datasets, 809 rows, 7 demo logins` / `Seed loading complete {inserted:809, errored:0}`; after: `loading 17 datasets, 818 rows, 7 demo logins` / `Seed loading complete {inserted:818, errored:0}`",
    "counts_before": "platform_admin ats_offer 14 · quillstone 5 jobs / 27 applications / 2 offers / 3 members / 2 inquiries (pending_approval 1, sent 1) · harborline 5 / 31 / 2 / 3 / 2 (approved 1, draft 1)",
    "counts_after": "platform_admin ats_offer 23 (accepted 9) · quillstone 5 / 27 / 3 / 3 / 2 (accepted 1, pending_approval 1, sent 1) · harborline 5 / 31 / 4 / 3 / 2 (accepted 2, approved 1, draft 1)",
    "accepted_rows_as_employer": "quillstone: `Offer · Avery Lindqvist → Robotics Software Engineer` status accepted, start_date 2026-09-18, expires_at 2026-09-02, approved_by usr_ats_quillstone_admin; harborline: `Offer · Elena Petrova → Welder` (start 2026-09-03, expires 2026-08-13) and `Offer · Karin Lindgren → Welder` (start 2026-09-06, expires 2026-08-21), both accepted, approved_by usr_ats_harborline_admin; seeker candidate01 sees exactly her own accepted offer (1)",
    "employer_dashboard_before_after": "unchanged, as it must be — no widget reads ats_offer. Quillstone open_jobs 3 / awaiting_action 18 / interviews_this_week 6 / pipeline 10-8-5-2-1(+1 rejected) before and after; Harborline 3 / 21 / 1 / 14-7-3-2-2(+3) before and after. Analytics via POST /api/v1/analytics/dataset/query with the widget filter as selection.runtimeFilter, each number cross-checked against a REST $filter count (all agreed), compiled SQL showed filter + row-level scope applied.",
    "cleanup": "both servers killed by their recorded pgid (24727, 27346); port 4403 verified free after each; worktree removed without --force after node_modules removal"
  },
  "doc_consistency": {
    "boot_banner": "computed from data: 17 datasets, 818 rows, 7 demo logins",
    "README": "818 x3 (line 18, boot table x2); isolation sentence Quillstone 3 offers, Harborline 5 / 31 / 4 / 3 / 2 — matches the runtime read-back",
    "AGENTS.md": "818 x2 (#61)",
    "src/data/index.ts": "811 ATS/identity rows + 7 logins = 818 (#61)",
    "DESIGN.md §06": "ats_offer 23 · 14 on offer-stage (3 pending_approval) · 9 accepted, one per hired",
    "pipeline.ts header": "offers 23 · one per offer-stage application (14) and one accepted per hired application (9)",
    "docs/screenshots/README.md": "809 kept as the shot-time fact + paragraph: 818 today, offers only, nothing photographed moved",
    "CONTRIBUTING.md": "example claim now 3 offers",
    "all_agree": true
  },
  "screenshots_still_valid": {
    "value": true,
    "evidence": "stage counts unchanged in artifact and via the funnel/pipeline datasets: 88/46/28/14/9 marketplace-wide, Quillstone 10/8/5/2/1(+1); Platform Overview, kanban, calendar, Find Jobs read employer/job/application/candidate/interview — none changed; no shot reads ats_offer; nothing reshot"
  },
  "mcp_calls": "0 — issue bodies and comments read via unauthenticated REST; PR creation and this comment via token-authenticated REST; no MCP search (no new issue filed, so no dedup search needed)",
  "open_questions": [],
  "out_of_scope_findings": [],
  "findings": [
    "STALE COMMENT, NOT MINE TO EDIT: src/dashboards/employer-hiring.dashboard.ts still says 'the demo seed has no offer on any hired application (its 14 offers all sit on offer-stage applications)'. src/dashboards/ is on this card's ⛔ list; #55 builds that very tile and lists '#53's seed fix' inside its own scope, so it owns the rewrite. After #64 merges, #55's seeding bullet is already done and its body item 4 is stale — PM may want to trim #55.",
    "MEASUREMENT TRAP (upstream, not filed — PM to decide whether it is a defect or permissive-by-design): POST /api/v1/analytics/dataset/query accepts a filter at body.filter, selection.filter, filters or where with HTTP 200 and silently ignores it (compiled SQL carries only the row-level scope; Quillstone open_jobs reads 5 instead of 3). The key the analytics service reads is selection.runtimeFilter (the Console uses it). A probe that does not print the SQL would report unfiltered tenant counts as widget numbers.",
    "OBSERVATION, NOT A DEFECT: application a106 stays at stage `offer` with a `declined` offer — a legitimate transient (employer deciding whether to counter), pre-existing, allowed by the machine; left as is.",
    "TOOLING: /home/user/objectstack/scripts/pm/os-verify-lock.sh is mode 644 on this box, so an executable check skips it silently; `bash path -c` works. The final gate pass ran through it.",
    "PROCESS NOTE: PR title is conventional-commit style rather than '= issue title' (AGENTS.md): the issue title is a problem statement and two issues are folded; #62/#57/#54 used the same style. Footer: dispatch asked for the bare form, the os-dev rules for the session-URL form, the harness for its own block — the PR carries the session-URL rule-line block plus the harness lines; on read-back the platform moved the rule-line block to the very end, content intact. docs/evidence/issue-8 and docs/backlog/10-seed-data.md still say 14 offers / 809 rows: dated records, left alone."
  ]
}

---

Generated by Claude Code