docs: propose owner-only kas sign-in from private messages
Problem / Motivation
Goal: Obtain a reviewable maintainer decision on terminal-free, owner-only KAS account sign-in from Telegram and Slack private messages before implementation begins.
Part of kirodotdev/KiroCrew#26042
The upstream issue is tier:T3 and parked. This is the separate RFC requested by the maintainer at https://github.com/kirodotdev/KiroCrew/issues/26042#issuecomment-6514102354, not a claim, implementation assignment, or request to treat the feature as scheduled.
Why it matters
A phone-operated remote gateway needs an explicit recovery entry point that does not depend on a working authenticated model turn or a terminal on the host. Account sign-in changes the installation's credential, so sender identity, private-DM proof, cross-entry-point races, and disclosure must be reviewed before a command is added.
Not a goal
- Implementing or enabling a login command, changing runtime behavior, or changing the issue's priority or labels.
- Closing kirodotdev/KiroCrew#26042, recording design acceptance, or claiming live Telegram/Slack/provider validation.
- New credential stores, shell execution, owner enrollment through chat, account switching, shared-channel login, or enterprise boot-policy changes.
What changed (motivation → approach → change)
Added docs/request-for-change/rfc-kas-owner-dm-sign-in.md with the upstream front matter (status: draft), an explicit unimplemented status, and source references pinned to 8b9aeceeef98a03a491d3547f2b71632d06c1765.
The RFC maps Telegram commands/transport/session-resume ownership, Slack native and transport dispatch, configured owner checks, dashboard owner gating, KasLoginService, device authorization, the encrypted account vault, and current KAS route resolution. It distinguishes the merged CLI/cloud sign-in work (kirodotdev/KiroCrew#23657), sign-in generation (kirodotdev/KiroCrew#25608), and remote MCP sign-in retention (kirodotdev/KiroCrew#28630) from this account-login entry point.
The proposed direct-device-flow MVP has a fresh confirmation, immutable owner/installation/DM binding, one shared per-vault interactive attempt, fenced persistence, cancellation/timeout outcomes, replay controls, and secret-free history/audit requirements. The existing independently authenticated login UI is compared explicitly. Owner-binding authority and the shared UI/CLI coordination cost remain blocking maintainer decisions.
Linked the proposal from docs/README.md so it is reachable from the contributor documentation index. No source code, configuration defaults, workflow, shipped behavior specification, or changelog is changed.
Branch provenance: the preexisting knowlet:docs/rfc-kas-owner-dm branch had no unique changes (45 commits behind the audit snapshot) and was fast-forwarded from 037a89c872b839c9f83772d6dbc7043f9ce0289b to the audited upstream commit before this standalone documentation commit. No write performed for this work targeted the fork's main.
Publication boundary: this is a staging Draft PR in the fork, not an upstream submission. Creating the upstream Draft PR and posting a pointer on the upstream issue both returned 403 Resource not accessible by integration. The base branch docs/rfc-kas-owner-dm-base is an unchanged snapshot of the audited upstream commit, so this fork review contains only one design commit and two documentation files, not unrelated upstream changes. Do not merge this staging PR as an implementation. The intended upstream target remains kirodotdev/KiroCrew:main, with head knowlet:docs/rfc-kas-owner-dm.
Backwards compatibility
Compatible: documentation only; no command, API, credential, or accepted input changes.
Removes nothing: the RFC proposes future behavior and does not remove or tighten anything on main.
Tests
Performed 29 local document checks only:
- YAML front matter, draft status, issue linkage, required RFC sections, and the unstarted implementation list.
- Reference definitions and immutable source-reference pin consistency; source paths checked against the files actually read through GitHub.
- Upstream
docs/README.mdsource reconstructed and verified against its Git blob SHA84606a325a0ed90e153bbb02ec1b0a4d4b608711before the focused index addition. - Whitespace/conflict-marker and PR-template-heading/Goal checks on the authored files and prepared upstream description.
- GitHub compare confirms the design commit is one commit ahead of the audited base and changes only the RFC and
docs/README.md.
Not run: full-repository scripts/docs-lint.sh, scripts/local-gate.py, backend/frontend tests, and live OAuth/platform E2E. The execution container could not resolve/reach GitHub for a full checkout; source inspection and publication use the GitHub connector. This is not a claim that CI or the repository-wide gates passed. The RFC's acceptance matrix specifies future tests, not completed tests.
Remaining before merge-readiness: add the proposal row to the directory-level docs/request-for-change/README.md index and run the upstream repository documentation gate. The root index link is present, but is not represented as completing that separate directory-index requirement. This PR intentionally remains Draft for design discussion.
Proof: not provable -- docs or config only
Manual verification
Read the current issue and maintainer guidance, RFC format, contribution rules, PR template, and the pinned source integration points. Rechecked the merged state of kirodotdev/KiroCrew#23657 and kirodotdev/KiroCrew#25608. No provider device authorization was initiated, no account credentials were read, and no Telegram or Slack message was sent.
Requested feedback: authoritative owner binding; direct code/link versus authenticated UI; shared single-flight/commit fencing; first supported providers and gateway shapes. Implementation remains blocked on that feedback and the normal acceptance process. No upstream review or successful upstream notification is claimed.
Screenshots / video
N/A — a design document only, with no rendered product change.
Related Issues
Part of kirodotdev/KiroCrew#26042
Related existing work: kirodotdev/KiroCrew#23657, kirodotdev/KiroCrew#25608. Distinct remote MCP credential lifecycle: kirodotdev/KiroCrew#28630.
Sibling sweep
N/A — this is a standalone design proposal, not a fix or revert.
Pattern harvest
N/A — no production defect is fixed in this documentation PR.
Checklist
- At most two commits (one is the norm), with a Conventional Commits title (
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...) - Existing tests pass and new tests added for new functionality — not claimed; no runtime functionality is implemented here.
- Self-review completed; code follows project style guidelines — documentation and source-backed design review only.
- Documentation updated (if applicable) — RFC and root index added; directory-level index and full documentation gate remain outstanding.
- No secrets, credentials, or internal references in the diff
Contribution License Agreement
Contributed under the project's Apache-2.0 license.
Summary by Sourcery
Document and solicit maintainer decisions on a secure, terminal-free owner-only KAS sign-in flow without changing runtime behavior.
Enhancements:
- Add a draft RFC defining security, ownership, concurrency, lifecycle, and secret-handling requirements for future owner-only KAS sign-in from Telegram and Slack private messages.
Documentation:
- Link the new owner-only KAS sign-in RFC from the documentation index.
Tests:
- Add documented acceptance criteria and verification requirements for the proposed sign-in flow; no runtime tests are introduced.
Summary by cubic
Adds a draft RFC proposing terminal-free, owner-only KAS account sign-in from Telegram and Slack private messages, and links it from docs/README.md. This is documentation only; no runtime behavior, configuration, or credentials change.
The RFC maps the existing Telegram/Slack dispatch, owner checks, KasLoginService, device flow, and encrypted vault code to a proposed opt-in /login (!login) command grammar with fresh confirmation, replay defense, and fenced single-flight coordination, and explicitly marks owner-binding authority, direct-code delivery vs. authenticated UI, and shared commit fencing as blocking maintainer decisions.