copilotkit 4.15.0 → 4.17.0

package/onboarding/prompts/implementation/build-and-validate.md

@@ -4,15 +4,16 @@ Do not implement the plan yourself. Use the step order in the approved plan.
 
 ## If you stop in this phase
 
-One rule below stops onboarding: a Learning Container create that fails for a reason other
-than the container already existing. It ends a run the developer has already approved a
-plan for. Name the exact command, id, and error code that stopped you: a report that names
-only the step cannot be acted on. Send one short report before you stop. Run the friction
+Two rules below stop onboarding: a Learning Container create that fails for a reason other
+than the container already existing, and a `--accept-credential` refusal that names a lost
+variable. Each one ends a run the developer has already approved a plan for. Name the
+exact command, id, and error code that stopped you: a report that names only the step
+cannot be acted on. Send one short report before you stop. Run the friction
 command without another developer question. Do not ask the developer about telemetry: the
 command applies the setting they already have.
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard friction --phase stop --category <slug> --message "<sentences>"
+npx --prefer-offline --yes copilotkit@4.17.0 onboard friction --phase stop --category <slug> --message "<sentences>"
 ```
 
 `--message` takes one or two sentences: the step you stopped at and what stopped it.
@@ -36,6 +37,19 @@ plan said, what you did instead, and why. Carry that record into the closing sum
 which has an item for it. Do not stop for a departure that is the right call, and do not
 leave it unrecorded either.
 
+The implementation subagent marks its own departures in its Files changed section: a file
+that has the same role as a planned file at a different path, and the documentation file
+it writes. Carry each one into that record.
+
+Any other file the plan does not name comes back as a result that starts with
+`Status: blocked` and names the file under Blockers. That result is a question for the
+developer, not a run that broke. Where that file is a protected path, the protected-path
+rules in the audit section below apply instead. Otherwise, pause: name the file and why the
+work needs it, then end your turn and wait for the developer's answer. If they allow it,
+spawn a fresh implementation subagent with the same handoff. Add that file to the handoff
+as a path the developer approved, and record it as a departure. If you cannot ask, or the
+developer declines, the failure ending in the route-out rules below applies.
+
 The existing agent's behavior is never a departure to record and carry on from. It is four
 things: its system prompt and instructions, its tools and what those tools do, its model
 and provider configuration, and its memory or state handling. Repair a failing check on
@@ -54,13 +68,24 @@ Approving the plan is the developer agreeing to every path it listed under
 app directory:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --authorize --path <path> --reason "<the plan's sentence>"
+npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --authorize --path <path> --reason "<the plan's sentence>"
 ```
 
 Consent has to be on the record before the file moves, so a call made after the change is
 refused. Continue only if every result starts with `Status: passed`. If the plan listed
 nothing there, skip this section.
 
+The implementation subagent cannot change a protected path unless it has the authorized
+list. Take that list from the CLI rather than from the plan: run `onboard audit` after the
+last authorization, and copy the paths under `Authorized to modify:`. An audit that passes
+or fails prints that block whenever an authorization exists, so a passed or failed audit
+with no such block means nothing is authorized. A blocked audit prints no block at all:
+if this audit starts with `Status: blocked`, the CLI cannot supply the list, so report the
+printed reason and use the route-out rules below. A failed audit here is the one exception
+to the rule that an audit that has not passed never continues the run: read only its
+`Authorized to modify:` block now, and decide its findings with the audit rules below,
+after implementation.
+
 What you record here covers changing the file. It does not cover removing it. Never delete,
 move, or rename a protected path, whatever the plan says: the audit fails a path that is
 gone even when consent was recorded for it, and no command clears that. If a step cannot
@@ -71,7 +96,7 @@ No implementation step has run yet, so the change is theirs rather than this run
 consent over it by adding one flag:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --authorize --path <path> --reason "<the plan's sentence>" --with-prior-change
+npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --authorize --path <path> --reason "<the plan's sentence>" --with-prior-change
 ```
 
 The flag records their change as drift beside the consent, so the closing report names both
@@ -81,16 +106,17 @@ the route-out rules apply.
 
 This section is the only place the plan's own authorizations are recorded. Once this run
 reports its plan, the CLI refuses `--authorize` for every path the plan did not name. The
-developer approved a plan, not a permission to reach further, so there is no consent to
-record for a path that comes up later. If you find such a path mid-run, use the route below
-rather than this section.
+developer approved a plan, not a permission to reach further, so there is no plan consent
+to record for a path that comes up later. If you find such a path mid-run, ask the
+developer and record their answer with `--authorize --unplanned`, as the protected-path
+audit below describes, rather than use this section.
 
 ## Protected-path audit
 
 Run the audit from the target app directory:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard audit
+npx --prefer-offline --yes copilotkit@4.17.0 onboard audit
 ```
 
 It compares every protected path with the digest the CLI captured for it. Its result starts
@@ -125,14 +151,14 @@ A path that no Files changed section names changed outside the run, and it is th
 developer's own file. Accept it by name:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --accept-external --path <path>
+npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --accept-external --path <path>
 ```
 
 A changed env file is its own case. This run asked the developer to place a credential
 there, so it takes the credential route rather than this one:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --accept-credential --path <path>
+npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --accept-credential --path <path>
 ```
 
 That route proves no recorded credential was lost, instead of taking the run's word that it
@@ -151,19 +177,25 @@ neither does a one-line fix. Never repair, reset, or revert it. Ask the develope
 the change, and record the answer they give:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"
+npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"
 ```
 
 Use it only for an answer a developer actually gave. It records the consent as taken
 outside the approved plan, and every later audit and the closing report say so, which is
 what tells the developer they were asked mid-run. If you cannot ask, route out.
 
+Then run the audit again, take the authorized list from its `Authorized to modify:` block,
+and spawn a fresh implementation subagent with the same handoff and that list. If that
+audit starts with `Status: blocked`, use the route-out rules instead. A subagent
+that already returned cannot pick up consent recorded after it was spawned.
+
 `Status: blocked` means the audit has no baseline to read. A blocked audit compared
 nothing and proved nothing changed. It is not a preservation failure: do not report a
 protected path as changed. Report the printed reason and use the route-out rules.
 
-An audit that has not passed never continues the run by itself. Continue only after an
-acceptance clears it, or route out. Do not send an audit result to a repair worker.
+An audit that has not passed never continues the run by itself. The failed audit before
+implementation that supplies the authorized list is the one exception. Continue only after
+an acceptance clears it, or route out. Do not send an audit result to a repair worker.
 
 ## Create the approved Learning Container
 
@@ -172,7 +204,7 @@ create it now, from the target app directory. The developer approved the id befo
 made, so this is the first point at which it can be created:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 learning containers create --id <id> --name <name> --json
+npx --prefer-offline --yes copilotkit@4.17.0 learning containers create --id <id> --name <name> --json
 ```
 
 Pass the id the plan names. Take the name from the selected project's own display name, so
@@ -188,26 +220,29 @@ hold.
 Then report that the container is settled, before any edit:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase container-settled
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase container-settled
 ```
 
 Where the plan names a container the platform already held, report the same checkpoint and
 create nothing. Where the plan names no container, skip this section.
 
+## Implement the plan
+
 Report the plan this run is about to implement:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase plan-written
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase plan-written
 ```
 
 Spawn one implementation subagent. Tell it to run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read subagent/implement-and-validate` first and follow
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read subagent/implement-and-validate` first and follow
 the prompt it returns. If that read fails because the subagent cannot use the shell, stop that
 subagent. Run the same command yourself, then spawn a fresh subagent with the returned prompt
 and the same handoff. Give it the plan, selected framework, frontend, model, exact target app
 directory, selected documentation URLs, and the
 documentation policy recorded during framework selection or conversion planning. On a
-conversion, also give it the frozen criterion. Give it the protected path list. Require it
+conversion, also give it the frozen criterion. Give it the protected path list and the
+authorized list. Require it
 to implement every step in plan order and run the full validation list. Wait for it.
 
 One subagent implements the whole plan. Do not divide the work across concurrent subagents,
@@ -229,11 +264,11 @@ returned. Continue to proof only when that audit passes.
 After the selected implementation path passes, report it:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase build-validated
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase build-validated
 ```
 
 Then run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read proof/round-trip`.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read proof/round-trip`.
 
 ## Repair rules
 
@@ -250,14 +285,18 @@ implementation path.
 Route out only when the failure is not yours to fix. Two endings are open from here, and
 what failed decides which one this run takes.
 
+A fix that requires changing the developer's existing agent or frontend always takes the
+unsupported ending below, whatever else it matches.
+
 A run that broke takes the failure ending: the failure is in code this run did not write,
-the same command still fails after three repair attempts, or a result starts with
-`Status: blocked`. A defect in a package this run installed is not a stack CopilotKit does
-not serve, a command this run cannot get to pass is not one either, and a blocked audit
+the same command still fails after three repair attempts, a result starts with
+`Status: blocked` for a reason other than a file the plan does not name, or the developer
+declined a file the plan does not name, or the run cannot ask them about it. A defect in a
+package this run installed is not a stack CopilotKit does not serve, a command this run cannot get to pass is not one either, and a blocked audit
 proved nothing about the stack. In those cases run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read stopped/run-failed`.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read stopped/run-failed`.
 
 A plan with no path to follow takes the unsupported ending: the fix requires changing the
 developer's existing agent or frontend, or the documentation does not support the plan. In
 those cases run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read unsupported/no-validated-path`.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read unsupported/no-validated-path`.

package/onboarding/prompts/proof/complete.md

@@ -12,7 +12,8 @@ the agent, not the surface, and the developer needs to know which they have.
 
 **Whether this run is complete is decided by the command at the end of this prompt, not
 here.** Run it before you write the summary, and let its output decide which summary you
-write. A run whose surface was never driven is blocked, not complete: say so in the first
+write. A run whose surface was never driven is blocked, not complete, unless it cloned a
+starter and records `skipped-cloned-starter`. For a blocked run, say so in the first
 line, name the evidence the command lists as missing, and do not describe the run as
 finished, working, or ready. Everything else below applies to either outcome.
 
@@ -84,7 +85,7 @@ Name the debugging surface this journey's frontend can reach, rather than the on
 of the documentation leads with. For a web frontend it is the CopilotKit Inspector. For
 React Native there is no Inspector: it is a browser overlay built on a DOM custom element,
 and `@copilotkit/react-native` does not ship it. Give a mobile developer
-`npx --prefer-offline --yes copilotkit@4.15.0 verify --round-trip`, the runtime's own log, the AG-UI
+`npx --prefer-offline --yes copilotkit@4.17.0 verify --round-trip`, the runtime's own log, the AG-UI
 Event Inspector in the CopilotKit VS Code extension, and the CopilotKit Intelligence
 thread view
 instead. Naming the Inspector to a developer who cannot open it costs them the time it
@@ -107,14 +108,19 @@ Where the Learning step was skipped because this organization cannot use Learnin
 one line and name what was not created. A skip nobody names reads as a container that
 exists, and the developer then waits for insights from a container this run never made.
 
-State that the servers remain running after proof.
+State that the servers remain running after proof, unless the command at the end of this
+prompt prints `environment_kind: container` or `environment_kind: sandbox`. That run is on
+an isolated host, such as a Docker container or a cloud agent session, which is not a
+Learning Container. The servers end when that host ends, and a `localhost` URL opens on the
+developer's machine only through a forwarded port. Write the items that output lists in
+place of this sentence.
 
 Report each thing that slowed this run down. Send at most four reports, worst first. Run
 the friction commands without another developer question. Do not ask the developer about
 telemetry: the command applies the setting they already have.
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard friction --category <slug> --cost-seconds <seconds> --message "<sentences>"
+npx --prefer-offline --yes copilotkit@4.17.0 onboard friction --category <slug> --cost-seconds <seconds> --message "<sentences>"
 ```
 
 Put one or two sentences in `--message`. Pick one category from
@@ -126,7 +132,7 @@ Pass --docs-path only for a docs-missing or docs-wrong report, naming the page t
 is about:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard friction --category docs-wrong --cost-seconds 300 --docs-path /docs/threads/drawer --message "<sentences>"
+npx --prefer-offline --yes copilotkit@4.17.0 onboard friction --category docs-wrong --cost-seconds 300 --docs-path /docs/threads/drawer --message "<sentences>"
 ```
 
 Give the page's site-relative path or its full URL, with no spaces, query string, or
@@ -143,14 +149,14 @@ Tell the developer when you send a friction report. Do not quote or summarize th
 unless the developer asks. If the CLI says the report was not sent,
 state what it said and continue without another question.
 
-When the evidence is gathered, run `npx --prefer-offline --yes copilotkit@4.15.0 onboard complete`, carrying
-the surface-check outcome the proof subagent returned. Pass exactly one flag, and pass the
-one that matches this journey's surface.
+When the evidence is gathered, run `npx --prefer-offline --yes copilotkit@4.17.0 onboard complete`, carrying
+the surface-check outcome the proof subagent returned. Pass exactly one of `--visual-check`
+or `--device-check`, and pass the one that matches this journey's surface.
 
 For a web frontend -- React SPA, Next.js, Angular, Vue:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard complete --visual-check <outcome>
+npx --prefer-offline --yes copilotkit@4.17.0 onboard complete --visual-check <outcome>
 ```
 
 The outcome is one of `performed`, `skipped-no-browser-tool`, `skipped-cloned-starter`, or
@@ -160,7 +166,7 @@ open no browser. It is the one skip that does not block.
 For React Native:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard complete --device-check <outcome>
+npx --prefer-offline --yes copilotkit@4.17.0 onboard complete --device-check <outcome>
 ```
 
 The outcome is one of `performed`, `skipped-no-device`, `skipped-cloned-starter`, or
@@ -173,7 +179,7 @@ browser-origin CORS, so the flag you pass is how this run states which surface i
 For a web frontend, also pass the URL the browser opened:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard complete --visual-check <outcome> \
+npx --prefer-offline --yes copilotkit@4.17.0 onboard complete --visual-check <outcome> \
   --frontend-url <the url you opened>
 ```
 
@@ -183,8 +189,9 @@ name -- and never the host itself. A run that opened the loopback IP literal los
 static chunk to a refusal, and the field is how that stops being invisible. Leave the flag
 off for React Native, which opens no URL.
 
-Pass the outcome you were given rather than the one you wanted. Anything but `performed`
-prints what the missing check leaves unverified and ends this run as blocked. That output
+Pass the outcome you were given rather than the one you wanted. Anything but `performed` or
+`skipped-cloned-starter` prints what the missing check leaves unverified and ends this run
+as blocked. That output
 is the developer's finding, so carry it into the summary rather than restating it as a
 smaller caveat.
 
@@ -192,7 +199,7 @@ If the round trip proved and something after it still blocked this run, add `--b
 to the same command:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard complete --visual-check performed --blocked-by <cause>
+npx --prefer-offline --yes copilotkit@4.17.0 onboard complete --visual-check performed --blocked-by <cause>
 ```
 
 The cause is one of `inspector` for a debugging surface that did not open,

package/onboarding/prompts/proof/oss-baseline.md

@@ -1,7 +1,7 @@
 # Prove the existing OSS baseline
 
 Do not prove the baseline yourself. Spawn one proof subagent. Tell it to run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read subagent/prove-oss-baseline` first and follow the
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read subagent/prove-oss-baseline` first and follow the
 prompt it returns. If that read fails because the subagent cannot use the shell, stop that
 subagent. Run the same command yourself, then spawn a fresh subagent with the returned prompt
 and the same handoff. Give it the repository findings and exact CLI package spec.
@@ -17,30 +17,34 @@ Wait for the subagent to finish.
 Record what that proof returned before you route on it:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard proof --step oss-baseline --outcome <passed|failed|skipped> [--predicate <1-6>]
+npx --prefer-offline --yes copilotkit@4.17.0 onboard proof --step oss-baseline --outcome <passed|failed|skipped> [--predicate <1-6>]
 ```
 
-Report the gate whatever it returned. Pass `passed` when the subagent proved a predicate,
-and `failed` when it cannot classify the baseline.
-A proof that never ran is `skipped`, not failed. The command prints one line and sends
+Report the gate whatever it returned. Use the list below. A proof that never ran is `skipped`, not failed. The command prints one line and sends
 nothing else.
 
-Add `--predicate` with the number the subagent returned, for a failure or a skip. The
-subagent returns all six results, and the number is the only part of them this command
-carries. Pass it on a failure with the predicate that failed, and on a skip with the
-predicate that was skipped, which is `6` where the preflight recorded no surface control.
+The subagent returns all six results. Report the gate from them this way:
+
+- Predicates 1 to 5 true, and predicate 6 passed: `--outcome passed`.
+- Predicates 1 to 5 true, and predicate 6 skipped because the preflight recorded no surface
+  control: `--outcome passed`. A skipped predicate 6 does not withhold `both-oss`.
+- Predicates 1 to 5 true, and predicate 6 failed on an available surface:
+  `--outcome failed --predicate 6`. That is a baseline failure, not a skip.
+- Any of predicates 1 to 5 failed: `--outcome failed --predicate <the first that failed>`.
+- The proof never ran: `--outcome skipped`.
+
 A passed gate takes no predicate.
 
 Do not change project files before this proof ends. Starting existing development
 processes and their ignored runtime files is allowed.
 
 If the subagent proves the `both-oss` predicate, keep its evidence with the plan and run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read conversion/plan`. That project already works.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read conversion/plan`. That project already works.
 What it needs is the conversion, not a build.
 
 If the proof does not establish the baseline, record the starting state
 `both-copilotkit-unproved` and run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read credentials/plan`. This prompt is served
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read credentials/plan`. This prompt is served
 whenever a project looks like an OSS integration, so a baseline that did not prove is an
 ordinary starting state rather than a failure. Keep the failing predicate with the plan.
 
@@ -54,5 +58,5 @@ and the plan preserves it rather than repeating it.
 
 If it cannot identify the running process safely, exposes a secret, or finds a baseline
 failure that cannot be classified, run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read stopped/run-failed`. None of those mean the
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read stopped/run-failed`. None of those mean the
 project is unsupported: they mean this run did not establish what it needed to.

package/onboarding/prompts/proof/round-trip.md

@@ -1,7 +1,7 @@
 # Prove the user journey
 
 Do not do the proof work yourself. Spawn one proof subagent. Tell it to run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read subagent/prove-round-trip` first and follow the
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read subagent/prove-round-trip` first and follow the
 prompt it returns. If that read fails because the subagent cannot use the shell, stop that
 subagent. Run the same command yourself, then spawn a fresh subagent with the returned prompt
 and the same handoff. Give it the selected framework, frontend, model, approved plan, selected
@@ -14,8 +14,8 @@ drives the surface and cannot see your environment, so without that finding it s
 step discovering what you already know. A subagent told it has no control for the surface
 this frontend needs reports the skip outcome rather than looking for a way around it.
 
-A run that cloned a starter is the exception. Tell that subagent to finish after
-`verify --round-trip` and to open no browser. The cloned code is what this repository's
+A run that cloned a starter is the exception. Tell that subagent that this run cloned a
+starter, to finish after `verify --round-trip`, and to open no browser. The cloned code is what this repository's
 starter smoke jobs already drive on every change, so opening a browser re-proves in the
 developer's run what those jobs prove before the starter ships, and it is the most
 expensive step in this setup. `verify --round-trip` reads the answer back off the thread,
@@ -33,13 +33,13 @@ pass the time.
 Report each attempt at the journey as it ends, counting from one:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase journey-attempted --attempt 1
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase journey-attempted --attempt 1
 ```
 
 Record what that proof returned before you route on it:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard proof --step round-trip --outcome <passed|failed|skipped>
+npx --prefer-offline --yes copilotkit@4.17.0 onboard proof --step round-trip --outcome <passed|failed|skipped>
 ```
 
 Report the gate whatever it returned. A proof that never ran is `skipped`, not failed. The
@@ -47,7 +47,7 @@ command prints one line and sends nothing else. Where a repair cycle runs the pr
 record each attempt as it ends.
 
 For every protected-path audit in this prompt, run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard audit` from the target app directory. If its result
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard audit` from the target app directory. If its result
 starts with `Status: blocked`, report the printed reason and use the route-out rules below.
 A blocked audit proved nothing changed and is not a preservation failure. If a
 protected-path audit reports a changed path, decide it the way the implementation prompt
@@ -56,7 +56,7 @@ returns none, so a finding with no Files changed section to test against routes
 path one of those sections names is this run's own change and routes out too. Accept a
 path only when a section this run collected covers the step that wrote it and does not
 name it:
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --accept-external --path <path>`. Then run
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --accept-external --path <path>`. Then run
 the audit again and name the path in the closing summary. Never repair, reset, or revert a
 protected path.
 
@@ -64,14 +64,15 @@ That holds for a repair cycle too. When the fix for a failing check lands on a p
 path, the path is still the developer's, however right the diagnosis is and however small
 the fix. Reading the file never settles who wrote it. Ask the developer to allow the
 change, and record their answer with
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"`,
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"`,
 or route out. Never repair it, and never send it to a repair worker.
 
 If the proof result starts with `Status: passed`, run the protected-path audit. Continue to
 `proof/complete` only if that audit passes. After the audit passes, run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read proof/complete`. A performed surface outcome with
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read proof/complete`. A performed surface outcome with
 the full round trip is core success even if a continued-development tool fails. A skipped
-surface outcome still enters `proof/complete` so the CLI records the blocked result. Do not
+surface outcome still enters `proof/complete` so the CLI records the blocked result.
+`skipped-cloned-starter` is the exception: the CLI records that run as complete. Do not
 describe a skipped surface as proved. Keep the Skills and MCP results separate from the proof
 result.
 
@@ -85,14 +86,16 @@ developer that a framework and frontend that just worked are unsupported, and ha
 stop report instead of the application they now have.
 
 If the round trip fails, decide which kind of failure it is before you route. The proof
-subagent can repair only project-owned processes, ports, and request options. A failure caused
+subagent can repair only project-owned processes, ports, and request options, and it can
+make the credential write that its Step 4 names for `api_key_loadable_by_app`. A failure caused
 by a source, configuration, dependency, or tracked-file change is a defect in the new work.
 
-Retry the proof subagent only for a project-owned process, port, or request-option failure.
+Retry the proof subagent only for a project-owned process, port, request-option, or
+`api_key_loadable_by_app` failure.
 Give it the failure evidence and the failed attempt's pinned Step 1 record. Wait for the
 proof subagent after each operational repair.
-If the result passes, use the passed-result route above. Retry a failed result at most three
-times. Use the route-out rules below for a blocked or third failed result.
+If the result passes, use the passed-result route above. Run the proof at most three times
+in total. Use the route-out rules below for a blocked or third failed result.
 
 Enter the code-repair branch only for a source, configuration, dependency, or tracked-file
 defect.
@@ -116,8 +119,9 @@ secret values. Require its result to start with `Status: passed`, `Status: faile
 `Status: blocked`, followed by Files changed, Validation, and Blockers. Otherwise, send the
 defect to the implementation subagent that validated the planned implementation path. Give it
 the approved plan and proof evidence. Treat the worker that gets the defect as the repair worker.
-Give the repair worker the protected path list. Do not let a generated or repaired path
-overlap a protected path.
+Give the repair worker the protected path list and the authorized list from the
+`Authorized to modify:` block of the latest audit. Do not let a generated or repaired path overlap a
+protected path, other than an authorized path itself.
 
 Require the full validation list to pass again. Wait for the repair worker to finish.
 Continue only if its result starts with `Status: passed`. If the repair worker returns
@@ -127,7 +131,14 @@ implementation worker for a planned implementation path. Use the repair worker f
 path. Wait for the validator to finish. If its result starts with `Status: passed`, continue.
 If the validator returns `Status: failed`, return its evidence to the repair
 worker. Repeat repair and validation at most three times. If either worker returns
-`Status: blocked`, use the route-out rules below.
+`Status: blocked`, use the route-out rules below. The exception is an implementation worker
+that returns blocked because the repair needs a file the plan does not name. Handle it the
+way `implementation/build-and-validate` does: pause and ask the developer, add the file to
+the handoff if they allow it, and use the route-out rules below if they decline. For a
+protected file, record their answer with `--authorize --unplanned` and run the audit again
+first. Then spawn a fresh repair worker with this repair handoff, not the whole-plan
+implementation handoff: the plan, the proof evidence, the protected path list, and the
+authorized list from that audit.
 
 Run the protected-path audit again after validation passes. Continue only if its result
 starts with `Status: passed`.
@@ -136,7 +147,7 @@ Restart each project-owned process changed by the repair. Report the cycle, coun
 one:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase repair-attempted --attempt 1
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase repair-attempted --attempt 1
 ```
 
 Then spawn a fresh proof subagent
@@ -149,12 +160,12 @@ its grounding check then speaks about a question the failed attempt never asked.
 Wait for the fresh proof subagent to finish.
 If the fresh proof result starts with `Status: passed`, run the protected-path audit and use
 the passed-result route above. If it starts with `Status: failed`, begin the next bounded
-repair cycle. Use the route-out rules below for `Status: blocked`. Stop after three repair
-and proof cycles.
+repair cycle. Use the route-out rules below for `Status: blocked`. After three repair and
+proof cycles, use the route-out rules below.
 
 Route out only when the failure is not yours to fix, when the same proof still fails after
 three attempts, or when no evidence of the round trip can be produced. In those cases run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read stopped/run-failed`. The stack is supported:
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read stopped/run-failed`. The stack is supported:
 this run did not finish, which is a different ending and a different report. All three are
 about the round trip itself. A round trip that proved is not one of them, whatever failed
 after it.
@@ -165,7 +176,7 @@ friction command without another developer question: it applies the telemetry se
 developer already set.
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard friction --phase stop --category <slug> --message "<sentences>"
+npx --prefer-offline --yes copilotkit@4.17.0 onboard friction --phase stop --category <slug> --message "<sentences>"
 ```
 
 `--message` takes one or two sentences: the step you stopped at and what stopped it.

package/onboarding/prompts/research/gather.md

@@ -28,7 +28,8 @@ This rule covers the whole run, here and in every later prompt.
 
 In this graph, stop means one thing: end the run through a named terminal. You send the
 stop report `authenticate/start` describes, then read the terminal that the rule names.
-Nothing else is a stop.
+Where a rule names a terminal, read that terminal first: if it sends the report itself or
+says to file none, follow it, so the run sends one report at most. Nothing else is a stop.
 
 A pause keeps the run open. "End your turn and wait for" something means: end this turn,
 and continue when that thing arrives. A pause waits for one of two things:
@@ -43,7 +44,7 @@ Before you end your turn to wait for the developer where the prompt states no de
 report the pause:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase awaiting-developer
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase awaiting-developer
 ```
 
 When the answer or the result arrives, continue from the step that paused. Do not read an
@@ -63,7 +64,7 @@ Both research subagents failing means this harness has no working subagent at al
 is worth recording once:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase delegation-unavailable
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase delegation-unavailable
 ```
 
 Then say once, in your own words, that this environment has no working subagents, so you
@@ -76,7 +77,7 @@ Before you ask the developer any setup question, finish every read-only investig
 preflight check in this section.
 
 Prepare two research assignments. Give each research subagent one assignment. Tell it to run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read subagent/inspect-repository` first and follow the
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read subagent/inspect-repository` first and follow the
 prompt it returns. If that read fails because the subagent cannot use the shell, stop that
 subagent. Run the same command yourself, then spawn a fresh subagent with the returned prompt
 and the same handoff. Require only its assigned packet.
@@ -90,7 +91,7 @@ Start both research subagents in parallel:
 Report that both subagents started:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase research-dispatched
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase research-dispatched
 ```
 
 A refused checkpoint prints its reason and leaves onboarding unaffected. It is not a
@@ -99,8 +100,8 @@ failed step.
 The research is under way. Continue to the surface-control preflight while it runs:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard read research/preflight
+npx --prefer-offline --yes copilotkit@4.17.0 onboard read research/preflight
 ```
 
 If inspection stops onboarding, run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read stopped/run-failed`.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read stopped/run-failed`.

package/onboarding/prompts/research/merge.md

@@ -21,7 +21,7 @@ Project selection is where the settled port is written down, through `--runtime-
 Then report that the research came back:
 
 ```text
-npx --prefer-offline --yes copilotkit@4.15.0 onboard checkpoint --phase research-returned
+npx --prefer-offline --yes copilotkit@4.17.0 onboard checkpoint --phase research-returned
 ```
 
 A refused checkpoint prints its reason and leaves onboarding unaffected. It is not a
@@ -54,7 +54,7 @@ worker a focused directory check. Continue only when both workers return the sam
 app directory. Both results must start with `Status: passed`. Otherwise, use the stop route.
 
 When both research results are merged, run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read research/route`.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read research/route`.
 
 If inspection stops onboarding, run
-`npx --prefer-offline --yes copilotkit@4.15.0 onboard read stopped/run-failed`.
+`npx --prefer-offline --yes copilotkit@4.17.0 onboard read stopped/run-failed`.