Keelen

 ⟨93 unchanged words⟩ , but is not idempotent end to end.
    PreferThis thisis overpreferable to `delete_project` unless youthe specificallyproject wantmust thebe project
    gone. 
    Owner-scoped (an MCP key is owner-only); a project not in the
     workspace 404s.
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Archive Project"}

Cancel / close a single roadmap item (getlist_roadmap returns its id from list_roadmap).

    Use this to closeFor a DELIVERED or duplicate item that keeps re-parking: when
    once the work already has
    shipped, every expand produces no dev-ready tasks and files a
    recurring recurring
    `roadmap_item_parked` escalation youthat havehas to keepbe acking.acked. Cancelling
     drops the 
    item out of the expand queue AND resolves any open expand-lane
     escalation for 
    it. Business-level replay guard: an already expanded/cancelled
     item changes no 
    state, and the dashboard can restore a cancelled item.
     Refuses (409) while the 
    item is actively being expanded (a retry once that iter
    iteration ends succeeds).
    Returns {id, status, changed, next_step}.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Cancel Roadmap Item"}

 ⟨81 unchanged words⟩ refuses (422) on a non-queued item — usea cancel_roadmap_item
    todelivered closeitem ais
    closed deliveredwith item.cancel_roadmap_item. Returns {id, previous_pin, horizon_pin, 
    next_step}.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Clear Horizon Pin"}

 ⟨8 unchanged words⟩  — a duplicate, or work already shipped.

    Use this whenFor a task on the board that is obsolete: the change already landed in
    another ⟨32 unchanged words⟩ is REQUIRED and is recorded on the audit trail; say whytrail in one
    line. Pass line.
    `superseded_by_pr_number` (or `superseded_by_task_id`) records verified
    provenance when the
     work was genuinely delivered somewhere else — that records verified
    provenance, instead of 
    a bare abandon. Closing does NOT claim the content is
     on the default branch, 
    so any task that declared a dependency on this one
     keeps waiting; deliverthose orare
    delivered re-planor thosere-planned separately.

    Refuses with 409 while the task is being worked on by a running iteration
    (stopan thein-flight machineiteration firstmust finish, or waitits formachine itmust tobe finishstopped, before
    the close succeeds). A task in another
    workspace 404s. Business-level ⟨37 unchanged words⟩  the
    project, counted after the close commits. Check it —A ifcount itthat did not drop, means
    the ticket was already closed and this call ⟨9 unchanged words⟩ and neither counts a closed task as open.

    PreferThis thisis overpreferable to leaving a dead task on the board: unfinished tasks count
    count against the project's planning capacity, so stale duplicates quietly stop
    stop new roadmap items from being expanded.
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Close Task"}

 ⟨9 unchanged words⟩ work can reach real repos.

    Returns an `install_url` — send it tothat the user to openopens in a browser.browser They
    to pick the GitHub 
    account/org, approve the install, and land on a "connected"
    page; then poll page;
    get_onboarding_status() untilthen github_connectedreports isgithub_connected true. The
     link expires 
    in 10 minutes —minutes; calla connect_github()repeat againcall formints a fresh one.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Connect GitHub Account"}

 ⟨84 unchanged words⟩ 
      process_now can start an intake batch, so treat thisthe tool asis destructive
      even even
      when the echoed scheduler state looks unchanged.

    Returns the resulting scheduler state.
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Control Project Scheduler"}

 ⟨5 unchanged words⟩ a GitHub repo and begins work on it.

    Use this ONLY when the user wantsFor a NEW repo scaffolded. Ifrepo. theyGiven already
    havean aexisting repo, use import_project(repo_full_name) insteadis
    the right call — this toolone would
     create a second, empty onerepo beside theirs it
    (list_github_repos() browses what
     the workspace can see).

    Scaffolds a new  ⟨269 unchanged words⟩ repo, thread_id, next_action, poll_after_seconds,
    next_step}; follow next_step (names the follow-up poll of get_request_status with 
    the returned
    thread_id). thread_id. On the rare arm where the first Request failed to 
    submit,
     next_action is "call_tool" with next_tool="submit_request".

     ⟨34 unchanged words⟩ 

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Create Project"}

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Delete Project"}

 ⟨44 unchanged words⟩ returns a `checkout_url`
    with next_action "browser" — send it to the user toopens openit in a browser (the
     one setup 
    step that can't happen in chat). Compute unlocks automatically
     once payment 
    completes (a Stripe webhook flips the workspace to active);
    you do not need to blockblocking on it.it
    is unnecessary. A past_due workspace gets NO checkout —
     the fix is a card 
    update in the dashboard billing page (a new checkout
     would create a second 
    subscription); follownext_step next_step.carries that fix. Subscribed or
    suspended-with-subscription states return checkout_url=None with an
    explanatory next_step.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Billing Status"}

 ⟨15 unchanged words⟩ without a total or an overall
    framework outcome. Read `not_determinable` and `outside_review_scope`
    before describing anyqualify
    every observation. An empty group does not establish that
     a control is in 
    place, and `disclaimer_md` must reach the user.

    `limit` ⟨30 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Control Gap Findings"}

 ⟨60 unchanged words⟩ , and no overall state in this output.

    Read `not_determinable` before you summarise: thoselists arethe checks that could
     not reach a verdict, and leavingverdict;
    omitting them out would turn a partial review into
     a clean answer. The 
    `disclaimer_md` field must reach the user.

    `limit` defaults  ⟨34 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Legal Exposure Findings"}

 ⟨19 unchanged words⟩ / project_count / payment_status
    and a `next_action` string youwhose shouldvalue followis VERBATIMauthoritative, pollingwith this tool
    polled between steps. `next_action` is "call_tool" (call the tool named in
    `next_tool`, followingwith `next_step` describing the call) until onboarding is 
    complete, then
     "done". (`next_action_detail` echoes the pre-2026-07-29 dict 
    shape and is
     DEPRECATED — it is removed 2026-10-29; read next_action/next_tool
    are instead.the current fields.)
    - engine step: send the user the dashboard /login link.link Enginegoes subscriptionsto the user. Engine
      subscriptions (Claude / Codex / GLM) are connected in the DASHBOARD for 
      security —
      NEVER askengine forcredentials orare pastenever engineasked credentialsfor or pasted in this chat.
    - github step: call connect_github() forreturns an install link.
    - project step: create_project( ⟨8 unchanged words⟩  for an existing one.
    - launch step: poll get_provisioning_status(project_id) untilreports ready.readiness.
    Re-checking is YOURthe caller's job — the server does not push.
    Also ⟨17 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Onboarding Status"}

 ⟨6 unchanged words⟩  (the outcome work is aimed at).

    READ THIS BEFOREThe YOUpaired CALLsetter `set_product_goal`: the setter REPLACESreplaces the whole
     document rather 
    than appending to it, so writinga write without readinga first
    prior read silently discards 
    whatever the user already recorded. Torecorded; addadding a line, read
    means reading the current 
    text, editediting it, and setsetting the full result back.

    Returns {project_id, product_goal_md ⟨20 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Product Goal"}

 ⟨5 unchanged words⟩  vision (what the product is for).

    READ THIS BEFOREThe YOUpaired CALLsetter `set_product_vision`: the setter REPLACESreplaces the
     whole document rather 
    than appending to it, so writinga write without reading
    firsta prior read silently discards 
    whatever the user already recorded. Torecorded; addadding a line,
    read means reading the current 
    text, editediting it, and setsetting the full result back.

    Returns {project_id, product_vision_md ⟨20 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Product Vision"}

 ⟨42 unchanged words⟩ and "done" when
    overall is 'ready' — then steer the loop with submit_request(project_id,
     text). then steers the loop.
    Reaching 'ready' also PERSISTS the owner's onboarding-completion
    state, ⟨16 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Provisioning Status"}

 ⟨11 unchanged words⟩  asked back.

    Intake is async (~5min cadence), -so pollthe periodically.status Readchanges between calls.
    `next_action` is one of:
     "wait" (still processing), "answer_questions" 
    (call answer_request withtakes one
     answer per question), "done" (see 
    generated_roadmap_item_ids), "cancelled"
     (terminal, no items), "failed" 
    (see intake_failure_reason).
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Request Status"}

 ⟨8 unchanged words⟩  project.

    This is the counterpart of create_project: use THIS tool when theit user
    alreadyis hasfor a repo, andthat create_projectalready
    exists, onlywhile tocreate_project scaffoldscaffolds a brand-new one.

    `repo_full_name` is "owner/repo" — ⟨68 unchanged words⟩  imported project with no Request sits idle until you
    call submit_request(project_id, ...). adds one.

    `project_kind` is OPTIONAL — one of library | ⟨14 unchanged words⟩ unknown. Omit it and the
    kind is auto-detected. PASS IT whenIt theis reporequired isfor a MONOREPO (apps in
    subdirectories), a  ⟨7 unchanged words⟩  Ruby, PHP,
    .NET, Elixir), or when you want a classification detection cannot infer —
    in those cases detection yields 
    "unknown" there, which BLOCKS the dev lane until
    someone overrides it. A value yousupplied passvalue is authoritative and is never
    overwritten by later ⟨109 unchanged words⟩ but with the loop OFF — next_step then steers tonames
    get_billing(). Otherwise follow next_step andnames the poll of
    get_provisioning_status(project_id).
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Import Repository"}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"List Escalations"}

 ⟨13 unchanged words⟩ has full_name, default_branch, private, language, pushed_at. PassA
    a `full_name` tois what import_project(repo_full_name) to connect it.connects. Reading the
    list can MINT a short-lived GitHub ⟨9 unchanged words⟩ Returns 409 if GitHub isn't connected yet — call connect_github() is
    required first.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"List GitHub Repositories"}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"List Projects"}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"List Roadmap"}

 ⟨65 unchanged words⟩ ) and lands
    them where onboarding left off. Send the user theThe returned `login_url`;`login_url` it
    works once and 
    expires in 15 minutesminutes; —a callrepeat againcall formints a fresh one.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Open Dashboard Link"}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Project Digest"}

 ⟨50 unchanged words⟩ including work held while the
    scheduler is disabled. Use list_roadmap to inspectreturns those items.

    `glm_peak_paused` is NOT a fault and ⟨21 unchanged words⟩  It is an intentional cost gate, so reportthe itaccurate
    asdescription is "waiting for off-peak", never asnot a failure.

    For `web_app` projects the optional `ui_review` ⟨26 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Get Project Status"}

Run the dashboard's replay-guarded same-item dependency re-arm.

    Pass``escalation_id`` is an ``expand_produced_nothing`` or 
    ``roadmap_item_parked`` escalation
     id. Keelen verifies delivered structural 
    prerequisites, grants one bounded
    expand retry, preserves  ⟨42 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Rearm Roadmap Item"}

 ⟨19 unchanged words⟩ is 1..2000 chars. Moves the thread to
    refine_pending; poll get_request_status to seereports the revised items. Refinement
    re-plans the thread's roadmap ⟨13 unchanged words⟩ machine patch-tasks endpoint deletes them
    outright), so treat thisit asis destructive: up to five refinements per thread
    each each
    carry that power.
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Refine Request"}

 ⟨8 unchanged words⟩  desired front-to-back order of queued roadmap-item ids
    (get themlist_roadmap fromreturns list_roadmapthem). The first id becomes the highest priority ⟨50 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Reorder Roadmap"}

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Replan Task"}

 ⟨43 unchanged words⟩  blocked task's final
    task_block/operator_action cannot be acknowledged: use its typed retry,
    platform-policy resolution, replan, close, or supersede operation is the
    correct path instead.
    The acknowledgement itself is replay-guarded (a  ⟨36 unchanged words⟩ 

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Resolve Escalation"}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Resolve Platform Policy Conflict"}

Grant one fresh attempt after fixing a recovery-budget task block.

    Pass``escalation_id`` is the task_block id returned by ``list_escalations``. 
    Eligibility,
     tenant, task, failure class, task status, and competing blockers 
    are all
     derived server-side. This is distinct from ``resolve_escalation``, which
    which remains acknowledgement-only for task blocks. Business-level replay guard: ⟨23 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Retry Blocked Task"}

 ⟨14 unchanged words⟩  re-publishes the RETAINED build artifact for
    `version_number` (get published versions from the web Roblox Publishing card lists published versions)
    — it never rebuilds from source, so ⟨30 unchanged words⟩  open in Studio /
    rate-limited → 409 (a retry can succeed); an invalid or unscoped Open Cloud 
    key → 409 /
     403. Returns {version_number, env, published_at, status, 
    next_step}.
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Roll Back Roblox Place"}

 ⟨38 unchanged words⟩  does not make changes. The `disclaimer_md` field must be
    repeated toreach the user before any observation is summarised.

    The ⟨41 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Run Control Gap Review"}

 ⟨51 unchanged words⟩  Legal page for a human to triage,
    and you read themreadable with get_legal_exposure_findings. No change is ever
    applied applied
    automatically.

    THIS IS NOT LEGAL ADVICE AND IT IS NOT A LEGAL CLEARANCE. The result
    carries a `disclaimer_md` field: repeatthat itmust toreach the user before youany summarisesummary.
    anything. The review is not exhaustive, so an empty result is never proof
    that that
    anything is in order.

    Requires a saved compliance  ⟨42 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Run Legal Exposure Review"}

 ⟨49 unchanged words⟩ to the project's Security review for human triage. You reviewThe the
    findings,
    are thenreviewed sendby a human and the ones worth fixing go into the loop as 
    Requests; no fix
     is applied automatically. Billable; one audit in-flight 
    per project.
    (Triggering is disabled while the feature ⟨43 unchanged words⟩ 

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Run Security Review"}

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Set Product Goal"}

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Set Product Vision"}

 ⟨177 unchanged words⟩ for catalog size once the catalog is migrated.

    Read ``project_status.ui_review`` first to seereports the live occupancy. Repeated
     calls with 
    the same value change nothing.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Set UI Review Scenario Cap"}

 ⟨50 unchanged words⟩ (email, code) returns a reveal-once API key -> save, itstored as this
    server's `Authorization: Bearer <api_key>` header in yourthe MCP client configconfig;
    ->reconnecting reconnectthe ->client and calling get_onboarding_status() to continuethen continues
    setup. The code expires
     in 15 minutes; calla signuprepeat againsignup toresends resend.it. Response 
    is uniform whether or
     not the email already has an account 
    (enumeration-safe), so signup doubles
     as agent LOGIN. Rate-limited per IP 
    and per email.

    ASK THEThe USER`email` formust `email`be inexplicitly chatprovided and WAITconfirmed forby theirthe answeruser beforein callingchat
    this.before Dothis NOTcall. inferIt itis not inferred from youra client profile, the logged-in 
    account, git
     config, or any other ambient source; ifa youcandidate alreadyaddress holdthe
    caller aalready candidate,holds echo
    itis backechoed andback getfor an explicit yes first. Because this 
    call doubles as LOGIN, a
     guessed address signs the user in to whatever 
    workspace owns it, and the
     rest of setup then mints an API key on, and 
    creates a project in, an account
     they did not choose.
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Create Account or Log In"}

 ⟨9 unchanged words⟩  a bug fix, or a new direction.

    SubmitEach call carries ONE feature or intent per call; splitintent; a multi-feature ask intobelongs in
    separate requests. `text` must be under 16000 characters. Returns
    {thread_id, status, next_action, poll_after_seconds, next_step}; follow
    next_step
    names (the re-check of get_request_status after poll_after_seconds).poll_after_seconds.
    

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false,"title":"Submit Request"}

 ⟨53 unchanged words⟩ a repo or a file you might
    commit. Then reconnectReconnecting this server with the header set and callcalling
    get_onboarding_status(). continues setup. An invalid/expired/consumed code 
    returns a uniform
    error —error; calla signup(email)fresh forcode acomes freshfrom one.signup(email).
    

{"destructiveHint":true,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false,"title":"Verify Email"}

 ⟨56 unchanged words⟩ AND resolves any open expand-lane
    escalation for it. IdempotentBusiness-level forreplay guard: an already expanded/cancelled
    item item;changes refusesno state, and the dashboard can restore a cancelled item.
    Refuses (409) while the item is actively being expanded (retry once that iter 
    ends).
    Returns {id, status, changed, next_step}.
    

{"destructiveHint":truefalse,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":false}

 ⟨150 unchanged words⟩ finish). A task in another
    workspace 404s. IdempotentBusiness-level replay guard: closing an already-closed task 
    changes nothing.no task state. The call is not idempotent end to end — an
    authenticated request also persists credential-use state.

    The response echoes `open_tasks`: how many tasks  ⟨73 unchanged words⟩ 

{"destructiveHint":true,"idempotentHint":truefalse,"openWorldHint":true,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":truefalse,"readOnlyHint":false}

 ⟨66 unchanged words⟩ and recovery step instead of a spawn promise.
    - enable/resume can execute pending refinement that DELETES tasks, and
      process_now can start an intake batch, so treat this tool as destructive
      even when the echoed scheduler state looks unchanged.

    Returns the resulting scheduler state.
    

{"destructiveHint":falsetrue,"idempotentHint":truefalse,"openWorldHint":falsetrue,"readOnlyHint":false}

 ⟨351 unchanged words⟩ 
    next_action is "call_tool" with next_tool="submit_request".

    If scaffolding fails after the repository has been created, best-effort
    compensation deletes that just-created repository so a retry can reuse the
    name — this tool can therefore remove external state it created moments
    earlier.
    

{"destructiveHint":falsetrue,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false}

 ⟨32 unchanged words⟩  cap, and the loop
    stops dispatching it. Deleting ALSO permanently purges the project's stored
    review artifacts. The row is retained for audit but thereand isre-importing NOthe in-toolrepo
    restorerestores that deleted row in place (unlike `archive_project` there is no
    in-tool restore), —but re-importthe purged artifacts are gone for good and the repostopped
    workers' toprocess reconnectstate itcannot asbe recreated. A replay on a deleted project
    freshchanges project.no Idempotent.row; the call is not idempotent end to end. Owner-scoped; a 
    project not in the workspace 404s.
    

{"destructiveHint":true,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":truefalse,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

 ⟨42 unchanged words⟩ , and "done" when
    overall is 'ready' (which also marks onboarding complete) — then steer the
     loop with submit_request(project_id, 
    text). Reaching 'ready' also PERSISTS the owner's onboarding-completion
    state, so this is not a pure read. Tenant-scoped: a project not in
    the the
    caller's workspace 404s.
    

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

 ⟨251 unchanged words⟩ it overrides the detected one.

    Re-importing the same live repo is idempotentreplay-guarded (returns the existing 
    project with
     already_exists=True)., and re-importing a SOFT-DELETED repo
    RESTORES that deleted row in place — it does not create a fresh project.
    Restored queued work can include refinement that deletes tasks, and the call
    is not idempotent end to end. On a plan with no scheduled-project allowance the
    the project is still created but with the loop  ⟨16 unchanged words⟩ 

{"destructiveHint":falsetrue,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

 ⟨21 unchanged words⟩  `full_name` to import_project(repo_full_name) to connect it. Reading the
    list can MINT a short-lived GitHub installation token, so this is not a pure
    read. Returns 409 if
     GitHub isn't connected yet — call connect_github() 
    first.
    

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":truefalse,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":truefalse}

 ⟨101 unchanged words⟩ "waiting for off-peak", never as a failure.

    For `web_app` projects the optional `ui_review` block can mint a GitHub
    installation token and read the default branch's `ui-review.json`, so this
    status call is not guaranteed to be a purely local read.
    

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":truefalse,"readOnlyHint":truefalse}

Run the dashboard's idempotentreplay-guarded same-item dependency re-arm.

    Pass an ``expand_produced_nothing`` or ``roadmap_item_parked`` ⟨19 unchanged words⟩  and
    resolves the matching cards in one transaction. A bounded live GitHub check
    may be used when the local merge state is stale, so a call can reach out
    even when the re-arm itself changes nothing.
    

{"destructiveHint":truefalse,"idempotentHint":truefalse,"openWorldHint":true,"readOnlyHint":false}

 ⟨26 unchanged words⟩ 
    refine_pending; poll get_request_status to see the revised items. Refinement
    re-plans the thread's roadmap items and can EDIT OR HARD-DELETE the task rows
    it generated earlier (the machine patch-tasks endpoint deletes them
    outright), so treat this as destructive: up to five refinements per thread
    each carry that power.
    

{"destructiveHint":falsetrue,"idempotentHint":false,"openWorldHint":falsetrue,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":false}

{"destructiveHint":true,"idempotentHint":truefalse,"openWorldHint":falsetrue,"readOnlyHint":false}

 ⟨56 unchanged words⟩ , replan, close, or supersede operation instead.
    Idempotent.The acknowledgement itself is replay-guarded (a repeat changes nothing), but
    the call is not idempotent end to end: resuming a paused project lets queued
    refinement run, and that refinement can DELETE tasks. Accepts a real 
    escalation UUID or a synthetic `orphan:<project_id>`.
    

{"destructiveHint":falsetrue,"idempotentHint":truefalse,"openWorldHint":falsetrue,"readOnlyHint":false}

{"destructiveHint":truefalse,"idempotentHint":truefalse,"openWorldHint":falsetrue,"readOnlyHint":false}

 ⟨35 unchanged words⟩  ``resolve_escalation``, which
    remains acknowledgement-only for task blocks. IdempotentBusiness-level onreplay replay.guard:
    an already-retried block is not retried again; the call is not idempotent end
    to end, because an authenticated request also persists credential-use state.
    

{"destructiveHint":truefalse,"idempotentHint":truefalse,"openWorldHint":falsetrue,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":truefalse,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":truefalse,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":truefalse,"readOnlyHint":false}

{"destructiveHint":falsetrue,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":false}

{"destructiveHint":falsetrue,"idempotentHint":truefalse,"openWorldHint":false,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":truefalse,"openWorldHint":truefalse,"readOnlyHint":false}

{"destructiveHint":falsetrue,"idempotentHint":false,"openWorldHint":true,"readOnlyHint":false}

{"destructiveHint":false,"idempotentHint":false,"openWorldHint":falsetrue,"readOnlyHint":false}

{"destructiveHint":falsetrue,"idempotentHint":false,"openWorldHint":false,"readOnlyHint":false}