Repository navigation
skills(objectstack-automation): the screen row does not teach that a paused screen's Next / Submit and step count are engine-served (follow-up owed by #26657) #26756
Description
Filing gate ①: a follow-up owed by work in flight. PR #26752 (Fixes #26657) declares Skills-docs: follow-up owed — skills/objectstack-automation/SKILL.md. It keeps skills/** (Tier H) out of the PR, so the PR stays under the seat's landing. Its contract review (6444906744) PASSed and judged the split correct. Filed by domain:spec seat 1 (#6017, session session_01WGufzfhobWEgUUrjz4rrbh). ⛔ Not a claim.
Blocked-by: #26752
Business reason: once #26752 lands, a paused flow screen carries an engine-served progress (ScreenProgressSpec { nextScreen: 'none' | 'possible' | 'certain'; step?; stepCount? }). The console's FlowRunner reads it, showing "Submit" on a last screen, "Next" otherwise, and "Step N of M" only when every path agrees. An agent authoring a flow from the skill is not told any of this. It may hand-roll button labels or step counters into screen fields, or try to author progress on a screen node, and nothing catches that.
Reader who acts: the skills lane (skills/** is Tier H).
Wanted
- One line on the screen row of
skills/objectstack-automation/SKILL.md. It should say that Next / Submit and the step count come from the engine's reading of the graph, and that they are not authored. A branch, a fault edge, a loop back, a non-screen pause or a subflow answerspossible, which shows "Next" with no count. - An eval prompt only if it fits under the file's ceiling.
- The docs half is already in feat(automation): a paused screen says whether another screen follows, and the flow runner reads Next or Submit and counts steps #26752 (
content/docs/automation/flows.mdx).
Dedupe: open issues grepped for ScreenProgressSpec and for objectstack-automation + screen: none besides #26657 itself.
Clause-②: no
Metadata-protocol: none
Activity
objectstack-fleet commentedon Oct 27, 2026
Claim: PM loop round 1
Session:
session_01ADHkmSrRBdu9MdC6fc33UuAccount:
os-steve(the seat's linked user asGET /useranswers it; the card's assignee)Branch:
claude/project-thread-whp6tx(the card thread's harness push branch, self-reported at 2026-10-27T16:43Z — the measured platform exception toclaude/issue-N-slug)Worktree:
objectstack-issue-26756Domain:
domain:skillsSeat:
domain:skills#1File surface:
skills/objectstack-automation/SKILL.md(thescreenrow: Next / Submit and the step count are engine-served off the graph, never authored; a branch, fault edge, loop back, non-screen pause or subflow answerspossible), its row inscripts/check-skills-token-ratchet.mjs(fold first, then raise at the measurement under the standing ruling 「A:先削重复,余额按实测提高」), an eval prompt only if it fits with no raise, and a patch changeset for@objectstack/skills; nothing else (stop on breach; explain in the report)Container & model:
XS,mode:thread,model: CONTRACT_REVIEW_TIER(dispatch-gates --tier: MANDATORY —skills/**, clause ①, no one-line exit),effort: high. Tier H (skills/**): draft PR; ⛔ the dev never readies, arms or enqueues it.Clause-②: no
Metadata-protocol: none
Responsibility:
n/a — not a defect card(a follow-up owed by landed spec work)Thread-read: none
Premise: holds (seat re-measure on
origin/main8b276385: PR #26752 merged (e438a55e); the screen rows at SKILL.md:48 and :91 say nothing about progress, Next, Submit or step counts)Serial constraints cleared: the Blocked-by PR #26752 merged; no open PR names
skills/objectstack-automation/**objectstack-fleet commentedon Oct 27, 2026
os-dev-report
{ "issue": 26756, "status": "done", "branch": "claude/project-thread-whp6tx", "pr": "https://github.com/objectstack-ai/objectstack/pull/26812", "head": "270df422512b9c5b9afda3a89f4666bc9d5eaa62", "session": "session_01ADHkmS7nPHShBb87HjHRf4", "premise_still_valid": true, "summary": "The `screen` node row of skills/objectstack-automation/SKILL.md now says Next / Submit and \"Step N of M\" are not authored: when the screen pauses the engine reads the graph and serves `progress` (`nextScreen: 'none' | 'possible' | 'certain'`; `step` / `stepCount` only when every path agrees), which alone labels the button and draws the counter; a `decision`, `fault` edge, loop back, non-screen pause (`approval`, `wait`) or `subflow` after a screen answers `possible` (Next, no count); `progress` in a node `config` is refused (ScreenConfigSchema is a strictObject); never hand-roll a label or counter into `fields`. Every key read off origin/main (contracts/automation-service.ts, api/automation-api.zod.ts, screen-nodes.ts readScreenProgress, FlowRunner.tsx). Paid under the standing ruling: three restatements folded first (-87 bytes of text, -2 of re-wrap whitespace, each with its surviving home named in the PR body), then the token ceiling re-pinned AT the measurement 6358 -> 6466 (25429 -> 25864 priced bytes, +524 / -89, +435 net; headroom was 3 bytes). Patch changeset for @objectstack/skills with both declaration lines. No eval prompt: evals/flows-triggers-approvals.json is at headroom 0 (2734 / 2734), so a prompt fits only with a raise the card excludes. Whole-file 452 -> 451 lines, 6358 -> 6466 tokens; whole-package (11 SKILL.md) 4606 -> 4605 lines, 56135 -> 56243 tokens.", "tests": "No test files in the diff (published skill text, a ratchet row, a changeset). node scripts/check-skills-token-ratchet.mjs on head 270df422: `skills/objectstack-automation/SKILL.md is 6466 tokens (ceiling 6466; headroom 0)`, `81 authored bundle file(s) within their ceilings`; --self-test exit 0. No ablation: no tested behaviour changed.", "gates": "Local subset on head 270df422 (worktree objectstack-issue-26756, merge base 8b276385): build closure `os-verify-lock.sh -c \"pnpm exec turbo run build --concurrency=2 --filter='./packages/**'\"` VERDICT command-exit 0 (75/75, 8m43s). `dispatch-gates --commands --repo objectstack-ai/objectstack` derived 63 commands; each ran serially with stdin closed, exit captured before any pipe: 63/63 exit 0 (incl. check:skills-token-ratchet + --self-test, check:skill-docs, check:empty-changeset, check:pm-skill-id-lint, check:doc-authoring, check:doc-formula-expressions, check:dual-build-cjs-loads, check:dts-closure, check:pm-dispatch-gates). `dispatch-gates --ran`: 63 derived, 63 run, 0 NOT-MEASURED, 0 UNRUN. --tier: MANDATORY (skills/**, clause ①, no one-line exit); changed lines 48. CI families (pnpm lint, type-check lanes, the farm): the PR's check-runs, not re-run locally; CI not awaited.", "dogfood_battery": "n/a — no `!` in the PR title and no (narrowing) arm", "mcp_calls": "4 — all reads: issue_read get #26756 (1), issue_read get_comments #26756 (3, waiting for the Claim). No MCP write.", "api_writes": "2 — git push origin claude/project-thread-whp6tx (not REST); relay with-fleet.sh --via dispatch pr_create -> #26812 (run 40548961043, 9641 bytes sent = stored, assignee os-steve); this os-dev-report comment via post-stamped.mjs --comment=26756 (relay). No labels sent: a changeset is included, so skip-changeset does not apply and the delta names no label.", "line_budget": "skills/objectstack-automation/SKILL.md 452 -> 451 lines; tokens 6358 -> 6466 (ceiling re-pinned at the measurement under 「A:先削重复,余额按实测提高」); package 4606 -> 4605 lines, 56135 -> 56243 tokens.", "deviations": [ "The platform-assigned branch claude/project-thread-whp6tx is not claude/issue-N-slug (the claim records the exception); worktree objectstack-issue-26756 created on it after `git checkout --detach` in the shared checkout; no empty-branch probe push.", "origin/main moved to 0f0b9f55 after the branch was cut; dispatch-gates reports none of the visible upstream commits touched what the derivation reads, so main was not merged (AGENTS.md §10 conditions not met).", "The harness asks for issue_write / update_pull_request after pr_create (assignee + reviewer): not done — the standing order forbids GitHub MCP writes; the relay set the assignee, and on a Tier H diff the relay requests the governed approvers itself." ], "files_changed": [ "skills/objectstack-automation/SKILL.md", "scripts/check-skills-token-ratchet.mjs", ".changeset/26756-automation-skill-screen-progress.md" ], "open_questions": [ { "question": "Which screen row carries the teaching: the node row (L91) or the flow-type row (L48)?", "options": ["A: the node row, where config.fields[] is declared (chosen)", "B: the flow-type row"], "recommendation": "A, because the hand-rolled label or counter would be written into fields on the node row; the flow-type row only names the kind." } ], "out_of_scope_findings": [ "carrier: none · noted, not filed — a future raise of evals/flows-triggers-approvals.json under the same ruling can add a prompt asserting an answer never spells a Submit / Next label or a step counter into screen fields." ] }objectstack-fleet commentedon Oct 27, 2026
ACCEPT → PR #26812 comment 6448715005 ·
domain:skills#1· 2026-10-27T17:23Z · Tier H; draft until a GOVERNED_APPROVERS approval; contract review requested from the director seat.objectstack-fleet commentedon Oct 27, 2026
Landed ·
domain:skills#1·session_01ADHkmSrRBdu9MdC6fc33Uu· 2026-10-27T23:22Z — PR #26812 merged through the queue (merge commite269d3b652) on theos-zhuangapproval, after the director's PASS 6448877682; thescreennode row now says Next / Submit and the step count are engine-served fromprogress, never authored, its ratchet row at the measurement. The card closed by itsFixesline;pm:dispatchedstripped in this act.