[objectstack-fleet]

Requested by Warren · project thread
Refs objectstack-ai/cloud#3985

Clause-②: no
Metadata-protocol: none

维护者速读

• 改了什么：deploy/licensing.mdx 的「AI governance」一条，从「每次 AI 工具调用都有审计」改为「产品内置 AI 的每次工具调用」，并补一句：外部 agent 通过 /mcp 调数据工具不写 sys_ai_audit_log，RLS / FLS 和字段脱敏照样生效（与 docs(security): drop the certification, AI-audit and audit-export over-claims #455 同句）。
• 为什么改：cloud docs/design/ee-feature-catalog.md L55 明确「不要说每次 AI 调用都有审计」；docs(security): drop the certification, AI-audit and audit-export over-claims #455 只修了 reference/security.mdx。
• 译文：按翻译流程第 3 步删掉 licensing.ja.mdx / licensing.zh-Hans.mdx，gen:zh-hant 随之删掉 zh-Hant；没有改任何译文文件。
• 不改：why.mdx:58（AI 改数据本身走引擎写 sys_audit_log，不算说多）。
• 同句排查：英文内容中没有其他「every AI tool call / every AI call is audited」的原样重复（命中见下）。
• 两分钟验证：git diff origin/main --stat，预期只有 licensing 的 1 处英文修改和 3 个译文删除。
• 需要你做的：席位复审后合并。

What changes

content/docs/deploy/licensing.mdx, the "AI governance" bullet (L73–76 on main):

• before: "An audit trail of every AI tool call (who asked the AI what, and what it touched), …"
• after: "An audit trail of every tool call the product's built-in AI makes (who asked the AI what, and what it touched), … The trail covers the product's built-in AI. An external agent calling data tools over /mcp writes no sys_ai_audit_log row; RLS / FLS and field masking still apply to it."

The last two sentences are #455's wording on reference/security.mdx, verbatim.

Translation step 3: licensing.ja.mdx and licensing.zh-Hans.mdx deleted; pnpm --filter @objectos/docs gen:zh-hant removed the stray licensing.zh-Hant.mdx. No locale file is modified.

Same-claim grep (English content, locale files excluded)

grep -rniE "every (AI|tool) (tool )?call|every AI call|AI call is audited|all AI (tool )?calls" plus a pass over every line with both "audit" and "AI":

Hit	Verdict
deploy/licensing.mdx:73 "every AI tool call"	Fixed here
reference/security.mdx:258 "every tool call the product's built-in AI makes"	Already fixed by #455
build/ai-builder.mdx:25 "Every tool call is checked against that person's permissions"	Permission check, not an audit claim; true for external AI too. Not a repeat
build/agents.mdx:179 "audit:ai:tool events (per tool call …)"	Describes the agent runtime's events, not "every AI call"; not an exact repeat, left alone
why.mdx:58 "every AI-driven change is itself an audit entry"	Out of scope per the card (record writes go through the engine's sys_audit_log)
editions/license.mdx:50, editions/compare.mdx:35 "AI governance (action audit, …)"	Feature name only, no "every" claim

Gates (local, exit codes)

Gate	Exit
node apps/docs/scripts/gen-zh-hant.mjs --check	0
node .github/scripts/check-translations.mjs	0
node .github/scripts/check-translation-ownership.mjs --actor os-warren --files <listing>	0 (TRANSLATION_BOT_LOGIN unset, not enforced)
node .github/scripts/check-translation-output.mjs	0
pnpm --filter @objectos/docs build	0
node .github/scripts/check-positioning.mjs	0
node .github/scripts/check-locale-surface.mjs	0

🤖 Generated with Claude Code

https://claude.ai/code/session_01JQwxbWhdi8aV1BgkCTQ9h6

[objectstack-fleet]

repo:cloud#2 · os-warren · session_01JQwxbWTyRFQwYN1hZXbCHL

维护者速读

• 改了什么：「Licensing」页的 AI 治理一条，「每次 AI 工具调用都有审计」改为「产品内置 AI 的每次工具调用」，并补一句：外部 agent 通过 /mcp 调数据工具不写 sys_ai_audit_log，RLS / FLS 和字段脱敏照样生效。按翻译流程第 3 步删除这页的 ja / zh-Hans 译文，zh-Hant 由脚本跟随删除。
• 为什么改：原句说多了。审计只覆盖产品内置 AI（cloud#3985，与 docs(security): drop the certification, AI-audit and audit-export over-claims #455 修 reference/security.mdx 是同一类问题）。
• 风险与回滚：只改一句英文，删除的译文会按流程重新生成。回滚即 revert。
• 先看：content/docs/deploy/licensing.mdx L73–80。
• 两分钟验证：读这一段，和 docs(security): drop the certification, AI-audit and audit-export over-claims #455 的措辞一致。
• 需要你做的：无，席位落地。
• 席位意见：ACCEPT。「内置 AI」在同一段出现两次，略显重复，不值得再改一轮。

席位核对

ACCEPT · head fd4c208 · repo:cloud#2 · TIER_DEFAULT seat review (docs only; not governed; +8 −849, 4 files, 3 of them locale deletions)

1. Shape: matches claim 6731300343's file surface. The English page is edited, its ja / zh-Hans siblings are deleted, and zh-Hant follows gen:zh-hant. No locale file is edited.
2. Premise: the old sentence claimed "every AI tool call", and cloud ee-feature-catalog.md L55 says an external agent over /mcp writes no sys_ai_audit_log row. The new wording is docs(security): drop the certification, AI-audit and audit-export over-claims #455's, verbatim in its second half. The grep found no other exact repeat; near-misses are listed in the PR body.
3. Gates: thread readings at fd4c208 are 7/7 exit 0 (report 6731457319, gates-3985.txt): the docs build, positioning, locale-surface and the four translation checks. The seat readies now and arms auto-merge when the required checks are green.
4. Accepted deviation: none. Nit: "built-in AI" is said twice in the bullet; accepted as is.
5. Out of scope: none. Skills-docs: n/a.
6. Landing: not governed. pr_ready now; automerge_enable after green. Card cloud#3985 closes on merge by the seat's landing note (cross-repo Refs).