Repository navigation
docs(decisions): propose that a Bot's chat carries the app tools its owner ticked it for - #37812
chenmingwei23 wants to merge 1 commit into
Conversation
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
GPT 6.1 Review — 🔴 changes requested (blocking)GPT 6.1 found at least one blocking issue that must be resolved before merging This comment is updated in place on each push. BLOCKING -- docs/decisions/2026-11-15-a-bots-chat-carries-the-app-tools-its-owner-ticked-it-for.md:23 -- Superseding entry requires maintainer evidence and override
[BLOCK-MERGE] d9c7929 Adjudication (Opus 5.5) — is blocking on each finding proportionate?No adjudicable findings were provided. F1 (fenced): the harm rung is governance ceiling, because a decision nobody made would enter an immutable ledger. I confirmed the conditions on disk:
Recovery path: none within the ledger, since entries are written once and are only replaced by a new superseding entry. This happens on the PR's ordinary merge path, not in a rare corner case, so the claim stands. False positive or not applicable? A repository writer can comment: |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: the PR adds one decision entry and its index row. The entry's Notes
[DESIGN-REVIEWED] d9c7929 Description read[DESCRIPTION-READ] 06b9e40366ade9cf954939e2a4619ee216d2f8b0bfad381b7badde810d629560 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — 🔴 changes requested (blocking)Reviewed This entry records a decision no maintainer has made yet. Its BLOCKING — docs/decisions/2026-11-15-a-bots-chat-carries-the-app-tools-its-owner-ticked-it-for.md:3 BLOCKING — docs/decisions/2026-11-15-a-bots-chat-carries-the-app-tools-its-owner-ticked-it-for.md:5 [BLOCK-MERGE] d9c7929 Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
86c4090 to
b55ce0e
Compare
|
@iamwhatever (you decided the 2026-11-04 conductors-stay-simple entry) and @buluoray: please review this proposal. If you agree, restate the decision in a comment on this PR, then post What it asks: behind the default-off Open question: |
b55ce0e to
a708ddd
Compare
|
Backend Tests (3.12, 7) and Backend Tests (Windows) (2) were red on main, not on this docs-only change: both failed only test_denied_commands_security_ssh_self.py::TestSshFloorNestingDepth::test_the_gate_refuses_a_nest_past_the_bound_without_an_exception, which failed the same way on main 5cffe12 (the gate's 32-deep substitution cap from #37875 answered before the ssh floor). Coverage Gate followed them. Update: the fix #37964 is merged (529da1e). This PR is rebased onto main 529da1e and pushed as d9c7929, so those checks rerun on the fixed main. |
…owner ticked it for
a708ddd to
d9c7929
Compare
Docs only. Adds one proposed entry to docs/decisions and its README row. Merge order: this PR first, then the implementation PR #37813. It builds on #32983 (merged), which added
permissions.botsand the per-app ticked Bots.Problem / Motivation
Goal: record the maintainer decision that lets a Bot's own chat carry the tools of an app its owner ticked it for, so #37813 has a decision to cite.
The 2026-11-02 entry says a Bot's chat (the conductor) runs the stock conductor spec and takes nothing from the Bot's file. That leaves no way for an app the owner trusts to give a Bot its tools in the Bot's own chat. #32983 lets an app see and wake the Bots the owner ticked for it, but a woken Bot still cannot call the app's tools from its chat.
Why it matters
An owner who ticks a Bot for an app expects that Bot to work with the app from its own chat, with its own memory. Today the only route is a worker, which costs a second turn and puts the Bot's words in the worker. #37813 changes that behaviour, and the decisions ledger is where a change to a recorded decision must be recorded first.
Not a goal
What changed (motivation -> approach -> change)
The entry narrows one sentence of the 2026-11-02 entry and one of the 2026-11-04 entry (decided by @iamwhatever: "a mate's conductor mounts only Crew's own MCP servers"), behind the default-off
bot_chat_app_toolsfeature preview: an app's MCP server joins one Bot's chat, built per session like the Bot's worker spec, only while the app is enabled, offers the server incontributes.botChatServers, holds the owner-approvedtoolsBot scope inpermissions.bots, and the owner ticked that Bot. Losing any of those starts the chat's next turn on a fresh conversation without the server. The shared conductor spec is still written from source.The 2026-11-04 entry's reason is that a conductor only analyzes and dispatches. The entry's
## Whyanswers it directly: a Bot's own chat is also where the owner talks to the Bot and where its memory lives, so an app tool used there answers in that chat; routing it through a worker adds a turn and splits the reply across two chats. The exception is kept to the ticked Bot and the preview, so the reason still holds for every other conductor.The
toolsscope extends the field #32983 added (a third scope next toreadandwake) instead of adding a second switch.Backwards compatibility
Compatible: docs only; nothing that works today changes.
Removes nothing: the diff adds one decision file and one README row.
Tests
N/A -- docs only. The decisions index and history checks run in CI.
Proof: not provable -- docs or config only
Manual verification
N/A -- docs only.
Related Issues
Implementation: #37813. Builds on #32983.
Open items before merge
Decided by:line is PENDING until maintainers restate it on this PR: @iamwhatever, who decided the 2026-11-04 entry this narrows, and @bolichen97 or @buluoray, who decided the 2026-11-02 entry. Both restatements are needed. The entry is then updated to name them, link those comments under Evidence, and take the restatement's date in its filename andDate:line, because entries are immutable after merge.toolsas a thirdpermissions.botsscope, or a field of its own? It is settled here before merge, since changing it later takes a new superseding entry.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributed under the project's Apache-2.0 license.