Skip to content

FIX-1551: the CLI runs as the app's own principal - #2170

Merged
jhoffner merged 10 commits into
mainfrom
fix/FIX-1551-cli-principal
Sep 24, 2026
Merged

jhoffner merged 10 commits into
mainfrom
fix/FIX-1551-cli-principal

Conversation

@jhoffner

@jhoffner jhoffner commented Sep 24, 2026 •

Copy link
Copy Markdown
Collaborator

The problem

fsdev run and fsdev chat never asked the app who a caller is. They always ran as cli-user in the framework's placeholder organization, whatever the app's own resolver said. So a terminal run wrote records the app's users could never read. Nothing organization-shaped, such as mara hiring a seat into kitchen-sink, could be verified from the terminal.

What this does

The CLI now asks the loaded app the same question its HTTP routes ask, through the same code:

  • An app whose resolver names an organization runs there. In kitchen-sink that is devuser in kitchen-sink, and mara's hire from the terminal now lands there (VG, run for real below).
  • An app with no resolver stays cli-user in the placeholder organization. That behaviour is unchanged.
  • Every refusal stops the run before anything is written, exits 2, and names the way through. This covers a resolver that wants a credential, a session owned by someone else, a pre-organization session, and a hired seat pinned to another owner.
  • --org (on run and chat) and --user (on run) let a developer name the identity locally. serve and dev don't get these flags.
  • source: "cli" is reserved. Only the engine's in-process entry point can produce it, and a transport adapter that stamps it is refused.

What's asked of you

Nothing new. D1, the reserved-cli lock-in, and the F1 and F2 answers were approved on #2158.

Still owed before merge: a docs-writer and docs-editor pass on the four apps/docs pages. They are DOCS.md's approved draft, published close to verbatim, and this PR leaves apps/docs/** alone from here on.

Parts worth reviewing closely

  • The reserved mark (packages/engine/src/transports/host/createInboundTransportHost.ts).
    • A module-private WeakSet (inProcessAsks) holds the marked contexts. Only resolveInProcessPrincipal adds to it, after freezing the context, its envelope, its metadata and its body.
    • The guard in createPrincipalResolution refuses source: "cli" on any unmarked context, before any resolver runs.
    • The host and the in-process entry both call that one function.
    • metadata.body has the HTTP action body's shape, { userId, input }, so a resolver that reads the body sees the same thing both ways (Codex P1).
  • Nothing written before a refusal (packages/cli/src/principal.ts and packages/cli/src/commands/run.ts). The identity step, which run and chat share, applies the engine's own pinRejectsCaller to registry.pinOf(flow). Then the session check runs on the record already read. It uses isOrgAttributed / UnattributedOrgError, then flow, org and user. Only after both does the seed get written.
    • The engine now exports pinRejectsCaller, InstancePinCaller, isOrgAttributed and UnattributedOrgError so the CLI reuses the rules rather than copying them. These are new exports; nothing was moved.
    • A mismatched -s now exits 2, where before runAction refused it with exit 1.
  • Chat, per turn (packages/cli/src/chat/loop.ts): flow guard → identity (resolver, flags, pin) → owner check (rotates the session on a mismatch) → turn.

Checks (red / green)

Check Green Red seen by
V1 cli-principal.integration.test.ts V1: command.principal = dev/acme-dev/resolver, the stored seeded session is acme-dev, the app router's GET sessions/v1 returns 200, there is one [fsdev] running as stderr line, stdout is all JSON. BR-2: discovery gives cli-user/__fsd_default_org__/development-default. Real kitchen-sink after PR-B, model-free: KITCHEN_SINK_TEST_MODE=1 STORE_TYPE=filesystem pnpm fsdev run chat-agent run -i '{"message":"hi","mode":"ask"}' -s fix1551-v1 --capture … exits 0 with [fsdev] running as devuser in organization kitchen-sink (from the app's resolver), and .fsdev/data/sessions/fix1551-v1.json has devuser / kitchen-sink Test file run against origin/main's run.ts: expected 'ok cli-user@__fsd_default_org__' to be 'ok dev@acme-dev'
V2 Every existing run-command and chat test passes. Only change: chat-turn.integration.test.ts passes the new required orgId to executeTurn (7 call fixtures, no assertions changed) —
V3 echo / admin / digest / placeholder / bodyorg: the CLI outcome equals POST /<flow>/actions/respond for a credential-less caller. bodyorg names its org from body.input.message, giving ok body-user@org-hi both ways Planted fallback to the placeholder on refusal: admin: expected 'ok cli-user@__fsd_default_org__' to be 'refused: Action request requires non-…'. Planted body without input (the Codex P1 case): expected 'refused: Action request requires non-…' to be 'ok body-user@org-hi'
V4 A refused admin run with -s and --seed-session leaves no session and no request records, exits 2, and names the flow and --org Planted "resolve after the seed write": expected { id: 'v4', flowKind: 'admin', … } to be undefined
V5 --org (a credential flow runs as alice@acme), --org alone gives from: flag, blank / whitespace / placeholder --org are refused with nothing written, --user alone gives bob@acme-dev, --user alone on admin is refused, and a flag-org session is invisible to the app router while the app's own org is readable Planted "name the app's own org": expected 200 not to be 200
V6 -s on a placeholder-org session, --seed-session on a same-org/different-user session, and --seed-session on another org's session: each refused, owner named, record byte-for-byte unchanged Planted "seed before the check": the stored record gained "tampered":true
V7 fsdev chat: the admin turn fails naming --org, the echo turn succeeds as dev@acme-dev, /status shows Org: acme-dev, piped exit is 1. --org names every turn's org. An invalid --org is refused at startup Planted "chat ignores the resolver": the refusal text is missing and the session lands in the placeholder org
V8 serve/dev reject --org and --user as commander.unknownOption, and run/chat accept them under the same parse. Widened fence: no file under engine/src/routes, engine/src/transports (minus host/ and the re-exporting barrel), or the mcp, scheduled, node, next, vercel, bullmq and voice-openai sources references resolveInProcessPrincipal. Every root is asserted to have been scanned. FIX-1442 body-org tests and the node bind-guard tests pass untouched Planted --org on serve: expected { ok: true, options: { org: 'x' } } to deeply equal { ok: false, … }. A reference planted in packages/mcp/src, then in engine/src/transports/http: expected [ Array(1) ] to deeply equal [] both times
V9 An adapter declaring custom-ws that resolves a source: "cli" context through the shared host is refused, and so is a spread copy of a real in-process context. The same app's in-process ask still gets local-dev@acme Guard dropped: expected 'ok {"userId":"local-dev","orgId":"acm…' to match /^refused: source "cli" is reserved…/
Pin (review 1) Seat acme-dev.desk, pinned to dev@acme-dev. fsdev run with --user alice, and again with --org acme --user dev: refused with exit 2, and the store holds no session, no request, no seed. The owner runs. fsdev chat --user alice: the turn fails and writes nothing Pin check dropped: expected { id: 'pin-run', …(11) } to be undefined (session and seed written), and chat printed Chatting with acme-dev.desk … with no refusal
Unattributed session (review 1, BP-030) A session with no orgId, with -s and --seed-session: refused with the engine's UnattributedOrgError text ("stored before an organization was required … persistence guide"), and the record is byte-for-byte unchanged Branch dropped: expected 'Session "legacy" belongs to user dev …' to contain 'stored before an organization was req…'
--user source (review 1) run echo --user bob: stderr says [fsdev] running as bob (named by --user) in organization acme-dev (from the app's resolver), and the capture's from is resolver (the organization's source, as documented on CliPrincipalSource) Planted "--user not named": expected [ Array(1) ] to deeply equal [ Array(1) ]
Frozen context (review 1) The resolver receives a frozen context, envelope, metadata and body, and writing envelope.flowKind throws TypeError Freeze removed: expected [ false, true, true, true ] to deeply equal [ true, true, true, true ]
VG PASS, real model. pnpm tsx goals/cli-principal/runs-in-the-apps-organization/run.mts on a fresh .fsdev/data. See below FAIL (expected) with the CLI's ask replaced by the old placeholder answer. See below

VG, run for real (after merging #2159)

Model: kitchen-sink's intent/chat → vercel/anthropic/claude-sonnet-5, through AI_GATEWAY_API_KEY, with FSDEV_DEFAULT_MODEL and FSDEV_INTENT_* unset.

> tsx ../../packages/cli/bin/fsdev.ts run support.mara run -i '{"message":"Hire support.patmufzb4j2, an agent seat that takes refunds."}' --capture /tmp/fsd-goal-cli-principal-…/run.json
[fsdev] running as devuser in organization kitchen-sink (from the app's resolver)
[flow-state] action execution completed {… "userId":"devuser", … "orgId":"kitchen-sink", … "output":"Done — hired **support.patmufzb4j2** as an `agent` seat … Its address is `kitchen-sink.support.patmufzb4j2`."}
PASS — principal {"userId":"devuser","orgId":"kitchen-sink","from":"resolver"}; roster row workforce/roster/support.patmufzb4j2 under kitchen-sink, none under __fsd_default_org__; read from …/apps/kitchen-sink/.fsdev/data with no model

Control, with the CLI's ask replaced by the old placeholder answer (same command):

[fsdev] running as cli-user in organization __fsd_default_org__ (no resolver configured)
… "output":"The hire request was refused with this error: \"Organization id \"__fsd_default_org__\" must be lowercase letters, digits, and single hyphens …\""
FAIL —
  - command.principal is {"userId":"cli-user","orgId":"__fsd_default_org__","from":"development-default"}, wanted {"userId":"devuser","orgId":"kitchen-sink","from":"resolver"}
  - no roster row workforce/roster/support.patmufzc24i under kitchen-sink (found ["workforce/roster/support.patmufzb4j2"])

The first VG attempt refused to start. That was the .fsdev/data my earlier pre-PR-B V1 run had written, and PR-B refuses such a store by design ("Delete the store and restart"). I deleted it and re-ran. Both verdicts are in the goal's verdict log.

Suites run (merged tree c21ca3ba2; head c7088ab71 only adds the goal verdict log and the kitchen-sink README)

  • packages/engine, pnpm test: 212 files, 2703 passed, 4 skipped
  • packages/cli, pnpm test: 22 files, 273 passed
  • @flow-state-dev/kitchen-sink, pnpm test: 21 files, 152 passed.
    • The first run, straight after pnpm install, timed out one test at vitest's 5s default: named-org.test.ts › "binds the assistant's flow, a seat and a channel to kitchen-sink".
    • That file passed alone (the test took 3.5s), and a full re-run passed 152/152. It looks like cold-start transform time, not a regression.
  • @flow-state-dev/node, pnpm test: 5 files, 46 passed (bind guard)
  • Root pnpm typecheck: 70/70 tasks, plus the boundary, updater-purity, side-chain and control-shape guards
  • pnpm lint: 24/24
  • pnpm knip:ci: exit 0

origin/main has moved since the merge at 5f48575c0 and hasn't been re-merged. The new commits are:

Of these, FIX-1543's approved spec ("one owner-pin gate for hired seats") may later want to move the CLI's pin check onto the gate it defines. Today the CLI calls the engine's existing pinRejectsCaller.

Not verified here

  • The apps/docs editorial pass (docs-writer, then docs-editor) and a Docusaurus build or link check for #who-a-run-is. That work follows this PR's push.
  • FIX-1503: its verifier default is not on main, because createFlowState and the router still default to defaultBodyUserIdPrincipalResolver, so nothing needed reconciling. Whichever lands second must keep BR-2: a flow on the framework default answers the CLI with development-default.
  • FIX-1548 merged as fix(kitchen-sink): hired seats answer the admin credential that hired them (FIX-1548) #2148. It touched kitchen-sink (lib/workforce-admin-auth.ts, lib/workforce-registrar.ts, a test, the README) and docs, and no engine code. The CLI calls the host's own resolution, so its answer still matches HTTP.
  • The retained POC's characterization legs P1 and P8 describe the old behaviour, so they are expected to fail against this branch.
Contract
  • Surfaces S1–S7 per specs/issues/FIX-1551/PLAN.md. S1's shape: FlowState.resolveInProcessPrincipal(question) and resolveInProcessPrincipal({ registry, resolvePrincipal? }, question) → { userId, orgId, from: "resolver" | "development-default" }.
  • Pinned names: --org / --user, source: "cli", command.principal { userId, orgId, from } with from ∈ resolver | flag | development-default (the organization's source).
  • Changeset: minor for @flow-state-dev/engine and @flow-state-dev/fsdev, one sentence naming FIX-1551.
  • Guardrails: no copy of the resolution or pin rules outside the engine; the mark never leaves the host module; no route, transport or adapter references S1; no fallback to the placeholder when a configured resolver refuses.

Links: FIX-1551 · epic FIX-1455 · spec #2158 · VG dependency #2159 (merged)

🤖 Generated with Claude Code

https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB


Note

Medium Risk
Changes authentication and org boundaries for all CLI runs and touches shared inbound principal resolution; mistakes could write data under the wrong org or weaken the reserved cli source guard.

Overview
fsdev run and fsdev chat now resolve identity through the app, not a fixed cli-user in the development org. The engine adds FlowState.resolveInProcessPrincipal / resolveInProcessPrincipal, sharing the inbound host’s principal-resolution path with source: "cli" reserved (only in-process entry; transports that fake it are refused before resolvers run).

The CLI wires run, chat, and shared principal helpers: ask the app per invocation or per chat turn, honor --org / --user, enforce seat pins and session ownership before any write, exit 2 on identity refusal (including mismatched --session), log [fsdev] running as …, and record command.principal in --capture. Chat /status shows org; wrong-owner sessions rotate instead of running.

Exports used by the CLI: pinRejectsCaller, isOrgAttributed, UnattributedOrgError. Docs, kitchen-sink tests/README, a VG goal, and broad integration tests document and lock the HTTP-parity behavior.

Reviewed by Cursor Bugbot for commit c556bc9. Configure here.

fsdev run and fsdev chat stop answering "who is this caller" themselves.
They ask the loaded app through the same resolution every inbound host
uses (per-flow resolver before the host's, the user fallback, the
organization rules), so a terminal run lands in the organization the
app's HTTP routes would give it.

engine (S1):
- createPrincipalResolution: the host's resolution, extracted so the
  in-process entry and every inbound host call the same code.
- resolveInProcessPrincipal + FlowState.resolveInProcessPrincipal: the
  terminal's question (source "cli", no request, caller-named user),
  returning { userId, orgId, from }.
- source "cli" is reserved: a module-private WeakSet marks the contexts
  the in-process entry builds; the resolution refuses "cli" on any other
  context, whatever source the adapter declared.

fsdev (S2-S5):
- one identity step shared by run and chat; --org on run/chat, --user on
  run; refusal stops before any write, exits 2, names the flow and --org.
- run: an existing --session/--seed-session is checked for flow, org and
  user before the seed is written; the seed and the run use one identity.
- --capture records command.principal; one stderr line names the org.
- chat asks per turn; a refused turn fails and the loop continues;
  /status shows the organization.
- the hard-coded placeholder organizations and cli-user literals are gone.

S6: goals/cli-principal/runs-in-the-apps-organization (VG, needs
FIX-1500 PR-B). S7: changeset (minor engine + fsdev), package READMEs,
inbound-transports architecture note, and the apps/docs operations in
DOCS.md, published from its draft (docs-writer/docs-editor not yet run).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
@vercel

vercel Bot commented Sep 24, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

3 Skipped Deployments
Project Deployment Actions Updated
flow-state-dev-docs Ignored Ignored Sep 24, 2026 9:06pm UTC
flow-state-dev-kitchen-sink Ignored Ignored Sep 24, 2026 9:06pm UTC
flow-state-dev-ui Ignored Ignored Preview Sep 24, 2026 9:06pm UTC

Request Review

@linear-code

linear-code Bot commented Sep 24, 2026

Copy link
Copy Markdown

FIX-1551

@cursor

cursor Bot commented Sep 24, 2026

Copy link
Copy Markdown
Contributor

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_08e221df-abbd-4f75-a0e7-72dec9f7f1e6)

@jhoffner
jhoffner marked this pull request as ready for review September 24, 2026 20:07
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Sep 24, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-09-24T20:13:23.433287Z b75c1ed Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b75c1ed4cb

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread docs/architecture/inbound-transports.md Resolved
…ipal (FIX-1551)

Review round 1:
- A seat pinned to an owner refuses a terminal identity outside its pin
  in the shared identity step (run and chat), by the engine's own
  pinRejectsCaller over registry.pinOf, before any session, request or
  seed is written. Exit 2, like the other identity refusals.
- checkSessionOwner takes the record the caller already read, and gives
  a pre-organization (unattributed) session the engine's
  UnattributedOrgError message instead of "organization (none)" (BP-030).
- The stderr line names --user as the user's source; `from` is
  documented as the organization's source (values unchanged).
- The in-process context is frozen (context, envelope, metadata, body)
  before it is marked.
- The no-caller fence covers engine/src/transports (minus host/ and the
  barrel) and the mcp, scheduled, node, next, vercel, bullmq and
  voice-openai sources.
- Engine exports pinRejectsCaller, InstancePinCaller, isOrgAttributed
  and UnattributedOrgError so the CLI reuses the rules.

Codex round 1:
- P1: metadata.body now carries { userId, input }, the shape of an HTTP
  action body, so a resolver reading body.input answers the terminal as
  it answers HTTP. V3 gains a body-reading flow.
- P2: the changeset is one sentence naming FIX-1551, with the -s exit
  code change.
- P2: inbound-transports.md says `cli` is a principal-resolution source
  only and is not recorded on requests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
Brings in FIX-1500 PR-B (#2159) so VG can run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
…DME: mara hires from the CLI

VG passed on c21ca3b with intent/chat -> vercel/anthropic/claude-sonnet-5.
The control (the CLI's ask replaced with the old placeholder answer) failed
as expected on the principal and the roster row. PR-B landed first, so the
mara paragraph is reconciled here per DOCS.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
…ding (FIX-1551)

--org: run in this organization and skip the app's resolver. --user: the
organization comes from the app's resolver unless --org is also given;
the default user is the resolver's, or cli-user with no resolver or with
--org. Drops the "local only" and "still decides the organization" phrasing
the docs editorial pass removed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
…chen-sink tokens (FIX-1563)

#2184 (FIX-1552) merged catalog cases written against the pre-#2159 acme/bravo
constants, which #2159 removed, so main fails kitchen-sink typecheck. The
operator's leg now lists with the kitchen-sink admin token; the other-org leg
uses the token whose entry was refused at boot.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB
…-sink org (FIX-1563)

The test landed with `acme:tok-acme` as its admin credential. The app now
refuses any token that does not name `kitchen-sink`, so the admin flow
ran without its resolver and both admin-seat cases failed on main.

The test now boots the app's real `fsdev.config.ts`, as named-org.test.ts
does: the real host resolver, registrar and admin flow, with the token
bound to `kitchen-sink`. The in-app hire goes through mara's own `hire`
tool rather than a hand-built registration.

Both config-booting files get a 30s test timeout. A cold import of the
app takes about 4.8s alone, and two of them in parallel workers outran
vitest's 5s default.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ANuyMVi3YS653fcghb9gtB

Copy link
Copy Markdown
Collaborator Author

Typecheck & Test was red on 705a830eb, and the cause isn't in this PR. The failing step was kitchen-sink's typecheck:

  • test/hired-seat-auth.test.ts(181,38): Cannot find name 'ACME'
  • test/hired-seat-auth.test.ts(215,37): Cannot find name 'BRAVO'

Those constants were removed by #2159. #2184 (FIX-1552) then merged catalog cases that still use them, so main fails the same way. #2170 doesn't touch that file.

The fix from #2191 is ported onto this branch, head c556bc901:

  • e676aff7a merges origin/main.
  • a1221f165 cherry-picks cb80cdfa2, which re-keys hired-seat-auth's catalog legs to kitchen-sink tokens.
  • c556bc901 cherry-picks 4beaeb768, also from fix(kitchen-sink): re-key the hired-seat visitor test to the kitchen-sink org (FIX-1563) #2191. With only cb80cdfa2, main's hired-seat-visitor-auth.test.ts still failed on this branch: expected 404 to be 401 and expected 404 to be 201. Its admin credential names acme, which the app now refuses. 4beaeb768 re-keys that test to kitchen-sink.

Both cherry-picks become no-ops once #2191 merges.

The same push aligns fsdev run / fsdev chat --help and packages/cli/README.md with the apps/docs wording (dee661e06).

Validated at c556bc901:

  • kitchen-sink tsc: exit 0
  • kitchen-sink tests: 158 passed
  • cli tests: 273 passed
  • root pnpm typecheck: 70/70
  • pnpm lint: 24/24

Generated by Claude Code

@cursor

cursor Bot commented Sep 24, 2026

Copy link
Copy Markdown
Contributor

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_61371b95-9149-4d66-98cb-02811acc295d)

@jhoffner jhoffner left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Architect call — #2170 (FIX-1551 impl)

Verdict: wait (code direction looks right; docs pass still owed per PR.)

CLI identity through the app’s resolver (HTTP parity), reserved in-process source: "cli" (WeakSet mark; adapters refused), pin/session checks before any write, exit 2 on refusal — matches approved #2158. Engine exports reused rather than copied.

Hold merge until the apps/docs editorial pass lands (or is explicitly waived by the owner). After that: merge-ready from Architect’s seat.

Not owner approval / not the merge gate.

— FSD Architect

Copy link
Copy Markdown
Collaborator Author

The apps/docs editorial pass has landed. docs-writer applied the editor's first review in 705a830 and its second review in b4d303a. Every claim was checked against real fsdev run output: exit code 2, the quoted refusal line and the command.principal shape. The docs build is clean and the #who-a-run-is anchors resolve. dee661e brings the CLI's own --help and README wording in line with it. The "docs pass still owed" line in the body is out of date.


Generated by Claude Code

@jhoffner
jhoffner merged commit 7fd9eb6 into main Sep 24, 2026
13 checks passed
@jhoffner
jhoffner deleted the fix/FIX-1551-cli-principal branch September 24, 2026 21:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants