copilotkit 4.24.0 → 4.25.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +75 -14
- package/cli-build-info.json +7 -7
- package/index.js +8544 -6674
- package/onboarding/index.json +2 -2
- package/onboarding/prompts/authenticate/start.md +50 -45
- package/onboarding/prompts/conversion/plan.md +3 -3
- package/onboarding/prompts/credentials/finalize-plan.md +30 -8
- package/onboarding/prompts/credentials/plan.md +115 -49
- package/onboarding/prompts/credentials/settle-credentials.md +7 -5
- package/onboarding/prompts/credentials/write-plan.md +15 -15
- package/onboarding/prompts/deployment/choose-local.md +44 -7
- package/onboarding/prompts/deployment/local-evaluation.md +12 -6
- package/onboarding/prompts/fallback/best-effort.md +8 -8
- package/onboarding/prompts/feature/a2ui/implement.md +15 -8
- package/onboarding/prompts/feature/a2ui/plan.md +15 -7
- package/onboarding/prompts/feature/a2ui/proof.md +14 -7
- package/onboarding/prompts/feature/a2ui/start.md +15 -14
- package/onboarding/prompts/feature/channels/implement.md +21 -9
- package/onboarding/prompts/feature/channels/plan.md +15 -7
- package/onboarding/prompts/feature/channels/proof.md +14 -7
- package/onboarding/prompts/feature/channels/start.md +20 -13
- package/onboarding/prompts/feature/chat-suggestions/implement.md +15 -8
- package/onboarding/prompts/feature/chat-suggestions/plan.md +15 -7
- package/onboarding/prompts/feature/chat-suggestions/proof.md +14 -7
- package/onboarding/prompts/feature/chat-suggestions/start.md +15 -14
- package/onboarding/prompts/feature/complete.md +2 -2
- package/onboarding/prompts/feature/learning/implement.md +32 -18
- package/onboarding/prompts/feature/learning/plan.md +15 -7
- package/onboarding/prompts/feature/learning/proof.md +37 -13
- package/onboarding/prompts/feature/learning/start.md +15 -14
- package/onboarding/prompts/feature/open-generative-ui/implement.md +15 -8
- package/onboarding/prompts/feature/open-generative-ui/plan.md +15 -7
- package/onboarding/prompts/feature/open-generative-ui/proof.md +14 -7
- package/onboarding/prompts/feature/open-generative-ui/start.md +15 -14
- package/onboarding/prompts/feature/realtime-sync/implement.md +20 -12
- package/onboarding/prompts/feature/realtime-sync/plan.md +15 -7
- package/onboarding/prompts/feature/realtime-sync/proof.md +14 -7
- package/onboarding/prompts/feature/realtime-sync/start.md +14 -13
- package/onboarding/prompts/feature/rich-threads/implement.md +21 -13
- package/onboarding/prompts/feature/rich-threads/plan.md +15 -7
- package/onboarding/prompts/feature/rich-threads/proof.md +14 -7
- package/onboarding/prompts/feature/rich-threads/start.md +14 -13
- package/onboarding/prompts/feature/stop.md +9 -3
- package/onboarding/prompts/feature/voice/implement.md +15 -8
- package/onboarding/prompts/feature/voice/plan.md +16 -8
- package/onboarding/prompts/feature/voice/proof.md +14 -7
- package/onboarding/prompts/feature/voice/start.md +15 -14
- package/onboarding/prompts/framework/ag2.md +6 -5
- package/onboarding/prompts/framework/agno.md +6 -5
- package/onboarding/prompts/framework/built-in.md +9 -6
- package/onboarding/prompts/framework/claude-sdk-python.md +2 -2
- package/onboarding/prompts/framework/claude-sdk-typescript.md +2 -2
- package/onboarding/prompts/framework/crewai-flows.md +12 -4
- package/onboarding/prompts/framework/custom-ag-ui.md +8 -2
- package/onboarding/prompts/framework/deep-agents.md +6 -5
- package/onboarding/prompts/framework/google-adk.md +2 -2
- package/onboarding/prompts/framework/langgraph-fastapi.md +8 -5
- package/onboarding/prompts/framework/langgraph-python.md +8 -5
- package/onboarding/prompts/framework/langgraph-typescript.md +8 -5
- package/onboarding/prompts/framework/llamaindex.md +6 -5
- package/onboarding/prompts/framework/mastra.md +9 -6
- package/onboarding/prompts/framework/ms-agent-dotnet.md +12 -4
- package/onboarding/prompts/framework/ms-agent-harness-dotnet.md +12 -4
- package/onboarding/prompts/framework/ms-agent-python.md +6 -5
- package/onboarding/prompts/framework/pydantic-ai.md +6 -5
- package/onboarding/prompts/framework/strands-python.md +10 -6
- package/onboarding/prompts/framework/strands-typescript.md +10 -6
- package/onboarding/prompts/frontend/angular.md +10 -3
- package/onboarding/prompts/frontend/nextjs.md +36 -3
- package/onboarding/prompts/frontend/plan.md +13 -7
- package/onboarding/prompts/frontend/react-native.md +9 -2
- package/onboarding/prompts/frontend/react-spa.md +9 -2
- package/onboarding/prompts/frontend/vue.md +9 -2
- package/onboarding/prompts/implementation/build-and-validate.md +45 -19
- package/onboarding/prompts/proof/complete.md +3 -3
- package/onboarding/prompts/proof/oss-baseline.md +30 -13
- package/onboarding/prompts/proof/round-trip.md +55 -17
- package/onboarding/prompts/research/gather.md +21 -11
- package/onboarding/prompts/research/merge.md +8 -2
- package/onboarding/prompts/research/preflight.md +2 -2
- package/onboarding/prompts/research/route.md +14 -7
- package/onboarding/prompts/starter/clone.md +73 -16
- package/onboarding/prompts/stopped/run-failed.md +8 -2
- package/onboarding/prompts/subagent/create-plan.md +9 -2
- package/onboarding/prompts/subagent/implement-and-validate.md +2 -2
- package/onboarding/prompts/subagent/inspect-repository.md +12 -6
- package/onboarding/prompts/subagent/prove-oss-baseline.md +6 -1
- package/onboarding/prompts/subagent/prove-round-trip.md +50 -9
- package/onboarding/prompts/unsupported/no-validated-path.md +8 -2
- package/package.json +1 -1
- package/release/release-tool.js +9 -3
|
@@ -9,17 +9,18 @@ two-context proof to subagents within the target project.
|
|
|
9
9
|
Run this before anything else in this prompt:
|
|
10
10
|
|
|
11
11
|
```text
|
|
12
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
12
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard identify --coding-agent <coding-agent-slug> --model <model-id> --model-family <model-family>
|
|
13
13
|
```
|
|
14
14
|
|
|
15
15
|
Report your product slug, such as `codex` or `claude-code`. The command needs no sign-in,
|
|
16
16
|
browser, or project. If this run already ran `onboard identify`, skip this step.
|
|
17
17
|
|
|
18
|
-
- `--model`: copy the model id exactly as your instructions write it.
|
|
19
|
-
|
|
20
|
-
- `--model-family`:
|
|
21
|
-
|
|
22
|
-
|
|
18
|
+
- `--model`: copy the model id exactly as your instructions write it. Without an exact
|
|
19
|
+
model id, leave out `--model` rather than guess.
|
|
20
|
+
- `--model-family`: give it even without `--model`. Closest of: `claude-opus`,
|
|
21
|
+
`claude-sonnet`, `claude-haiku`, `gpt-5`, `gemini`, `grok`, `glm`, `minimax`, `mimo`,
|
|
22
|
+
`deepseek`, `qwen`, `kimi`, else the model's series. Leave it out if you do not know it.
|
|
23
|
+
- Never give your product slug as a model, or as an unlisted family.
|
|
23
24
|
|
|
24
25
|
## Waits, stops, and subagents
|
|
25
26
|
|
|
@@ -37,13 +38,13 @@ carry on. Keep a read-only assignment read-only. If no subagent works at all, re
|
|
|
37
38
|
once:
|
|
38
39
|
|
|
39
40
|
```text
|
|
40
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
41
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase delegation-unavailable
|
|
41
42
|
```
|
|
42
43
|
|
|
43
44
|
Before you end your turn to wait for the developer, report the pause:
|
|
44
45
|
|
|
45
46
|
```text
|
|
46
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
47
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
47
48
|
```
|
|
48
49
|
|
|
49
50
|
A turn that ends while the run waits is a pause, not a stop, even when your harness ends it.
|
|
@@ -53,7 +54,7 @@ If you stop this run, send one short report before you stop. The friction comman
|
|
|
53
54
|
the telemetry setting the developer already chose, so it needs no separate question.
|
|
54
55
|
|
|
55
56
|
```text
|
|
56
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
57
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard friction --phase stop --category <slug> --message "<sentences>"
|
|
57
58
|
```
|
|
58
59
|
|
|
59
60
|
`--message` takes one or two sentences: the step you stopped at and what stopped it.
|
|
@@ -67,7 +68,7 @@ one scoped fix, and sends the report itself when no fix is approved.
|
|
|
67
68
|
Before any subagent or project process runs, capture the developer's existing work:
|
|
68
69
|
|
|
69
70
|
```text
|
|
70
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
71
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect
|
|
71
72
|
```
|
|
72
73
|
|
|
73
74
|
Keep the exact protected path list it prints and give that list to every subagent. No
|
|
@@ -82,20 +83,20 @@ round trip and inspect `/info`.
|
|
|
82
83
|
Wait for the inspection subagent to finish. Then report that the inspection came back:
|
|
83
84
|
|
|
84
85
|
```text
|
|
85
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
86
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase research-returned
|
|
86
87
|
```
|
|
87
88
|
|
|
88
89
|
If the inspection did not prove the baseline this intent extends, stop here without
|
|
89
90
|
changing files:
|
|
90
91
|
|
|
91
92
|
```text
|
|
92
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
93
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
93
94
|
```
|
|
94
95
|
|
|
95
96
|
When the inspection proved what this intent needs, read the next prompt now, in this turn:
|
|
96
97
|
|
|
97
98
|
```text
|
|
98
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
99
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/realtime-sync/plan
|
|
99
100
|
```
|
|
100
101
|
|
|
101
102
|
That prompt holds the plan, the developer's approval, and the paths the plan changes.
|
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
|
|
3
3
|
Delegate implementation to one subagent with the approved plan and fetched official guides.
|
|
4
4
|
If repository evidence does not prove an existing valid Intelligence selection, run
|
|
5
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
5
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 login --json` and follow its single streaming session. Then
|
|
6
6
|
list projects, show the developer the current/recent choices, and select or create one only
|
|
7
7
|
after they choose. Require the secret-safe selection summary to confirm the project file,
|
|
8
8
|
environment file, and key provisioning; never display a key. If existing configuration is
|
|
@@ -15,19 +15,20 @@ surface. Add the documented drawer only if the approved plan selected it.
|
|
|
15
15
|
|
|
16
16
|
An Intelligence runtime requires `identifyUser`, and it usually reads a signed-in session
|
|
17
17
|
that a CLI request does not carry. That makes the round trip UNKNOWN and the command exit
|
|
18
|
-
non-zero on an app that is working.
|
|
19
|
-
|
|
20
|
-
|
|
18
|
+
non-zero on an app that is working. It is not a defect to repair. The proof step proves
|
|
19
|
+
such an app on a thread the app created, with `--thread <id>`, which needs no session. Pass
|
|
20
|
+
`--header "Name: value"` only when the developer offers what `identifyUser` reads, or
|
|
21
|
+
record UNKNOWN as the expected result here.
|
|
21
22
|
|
|
22
23
|
Run focused type/test commands, then start the app and run
|
|
23
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
24
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 verify --expect-runtime intelligence --round-trip --agent <agent-id> --json`.
|
|
24
25
|
Repair changed-file failures before continuing. Record changed paths and secret-safe
|
|
25
26
|
validation evidence.
|
|
26
27
|
|
|
27
28
|
Before you spawn the implementation subagent, take the authorized list from the CLI:
|
|
28
29
|
|
|
29
30
|
```text
|
|
30
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
31
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard audit
|
|
31
32
|
```
|
|
32
33
|
|
|
33
34
|
From its result, copy the paths under `Authorized to modify:`. A passed or failed audit with
|
|
@@ -42,11 +43,18 @@ protected path, it must not edit it. It returns a result that starts with `Statu
|
|
|
42
43
|
and names the file under Blockers.
|
|
43
44
|
|
|
44
45
|
That result is a question for the developer. The file has not changed yet, so consent can
|
|
45
|
-
still go on the record. Name the file and why the work needs it
|
|
46
|
-
|
|
46
|
+
still go on the record. Name the file and why the work needs it.
|
|
47
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
47
48
|
|
|
48
49
|
```text
|
|
49
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
50
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
51
|
+
```
|
|
52
|
+
|
|
53
|
+
Then end your turn and wait for the developer's answer. If they allow it, record their
|
|
54
|
+
answer before any edit:
|
|
55
|
+
|
|
56
|
+
```text
|
|
57
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"
|
|
50
58
|
```
|
|
51
59
|
|
|
52
60
|
Continue only when the result starts with `Status: passed`. Run the audit again, take the
|
|
@@ -58,7 +66,7 @@ feature stop route below.
|
|
|
58
66
|
After validation and each repair, run:
|
|
59
67
|
|
|
60
68
|
```text
|
|
61
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
69
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard audit
|
|
62
70
|
```
|
|
63
71
|
|
|
64
72
|
Continue only when it starts with `Status: passed`. A path under `Authorized to modify:` is
|
|
@@ -69,7 +77,7 @@ with the implementation subagent's `Files changed` section. If that section does
|
|
|
69
77
|
the path, accept the developer's external change:
|
|
70
78
|
|
|
71
79
|
```text
|
|
72
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
80
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --accept-external --path <path>
|
|
73
81
|
```
|
|
74
82
|
|
|
75
83
|
For an env file where the developer placed a requested credential, use
|
|
@@ -82,11 +90,11 @@ Run the audit again after each accepted change. If it still fails, or starts
|
|
|
82
90
|
with `Status: blocked`, route out and stop:
|
|
83
91
|
|
|
84
92
|
```text
|
|
85
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
93
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
86
94
|
```
|
|
87
95
|
|
|
88
96
|
If validation cannot pass, or this intent needs a prerequisite the app does not have, use
|
|
89
97
|
the feature stop route above without further changes.
|
|
90
98
|
|
|
91
99
|
Otherwise run
|
|
92
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
100
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/rich-threads/proof`.
|
|
@@ -12,7 +12,7 @@ The approval is a wait for the developer. Before you end your turn to wait, repo
|
|
|
12
12
|
pause:
|
|
13
13
|
|
|
14
14
|
```text
|
|
15
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
15
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
16
16
|
```
|
|
17
17
|
|
|
18
18
|
A turn that ends while the run waits is a pause, not a stop, even when your harness ends it.
|
|
@@ -34,7 +34,7 @@ you plan. Run `onboard protect` again from the target app directory: on the same
|
|
|
34
34
|
prints the baseline the run already holds.
|
|
35
35
|
|
|
36
36
|
```text
|
|
37
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
37
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect
|
|
38
38
|
```
|
|
39
39
|
|
|
40
40
|
Then have one read-only subagent repeat the inspection for the findings above, and wait
|
|
@@ -59,19 +59,27 @@ The plan must name the documented Intelligence runtime wiring, server-side `iden
|
|
|
59
59
|
catch-all thread routes, the selected UI (drawer or existing headless UI), preserved behavior,
|
|
60
60
|
tests, and a reload/reopen proof. Ask for approval as its own question.
|
|
61
61
|
|
|
62
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
63
|
+
|
|
64
|
+
```text
|
|
65
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-plan-approval
|
|
66
|
+
```
|
|
67
|
+
|
|
62
68
|
The plan must list every protected path it needs to change under `Authorization requested`,
|
|
63
|
-
with one sentence explaining why. Write `None` when it needs none.
|
|
64
|
-
|
|
65
|
-
|
|
69
|
+
with one sentence explaining why. Write `None` when it needs none. A step that installs or
|
|
70
|
+
upgrades a package changes its manifest, so list a protected manifest such as `package.json`
|
|
71
|
+
or `pyproject.toml`. Its consent covers its lockfile, beside it or at the workspace root.
|
|
72
|
+
Authorization covers changing a protected path, not removing it, so the plan must not
|
|
73
|
+
delete, move, or rename one.
|
|
66
74
|
|
|
67
75
|
After approval, record each approved path before implementation:
|
|
68
76
|
|
|
69
77
|
```text
|
|
70
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
78
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --authorize --path <path> --reason "<the plan's sentence>"
|
|
71
79
|
```
|
|
72
80
|
|
|
73
81
|
If an approved path changed after capture and no implementation step has run, add
|
|
74
82
|
`--with-prior-change`. Continue only when every result starts with `Status: passed`. Do not
|
|
75
83
|
authorize a path the approved plan did not list.
|
|
76
84
|
|
|
77
|
-
Then run `npx --prefer-offline --yes copilotkit@4.
|
|
85
|
+
Then run `npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/rich-threads/implement`.
|
|
@@ -18,11 +18,18 @@ not edit it. It returns a result that starts with `Status: blocked` and names th
|
|
|
18
18
|
Blockers.
|
|
19
19
|
|
|
20
20
|
That result is a question for the developer. The file has not changed yet, so consent can
|
|
21
|
-
still go on the record. Name the file and why the fix needs it
|
|
22
|
-
|
|
21
|
+
still go on the record. Name the file and why the fix needs it.
|
|
22
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
23
23
|
|
|
24
24
|
```text
|
|
25
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
25
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
26
|
+
```
|
|
27
|
+
|
|
28
|
+
Then end your turn and wait for the developer's answer. If they allow it, record their
|
|
29
|
+
answer before any edit:
|
|
30
|
+
|
|
31
|
+
```text
|
|
32
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"
|
|
26
33
|
```
|
|
27
34
|
|
|
28
35
|
Continue only when the result starts with `Status: passed`. Run the audit again, take the
|
|
@@ -35,12 +42,12 @@ Report the outcome of each attempt as it ends, before any repair. The command co
|
|
|
35
42
|
attempts and the repairs:
|
|
36
43
|
|
|
37
44
|
```text
|
|
38
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
45
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard proof --step round-trip --outcome <passed|failed|skipped>
|
|
39
46
|
```
|
|
40
47
|
|
|
41
48
|
Use `passed` only for proved persistence and reopen, `failed` for an attempted proof that
|
|
42
49
|
failed, and `skipped` when the proof could not run. Then run
|
|
43
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
50
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 onboard audit`. Continue only when it starts with
|
|
44
51
|
`Status: passed`.
|
|
45
52
|
|
|
46
53
|
If the audit fails, never repair, reset, or revert a protected path. Compare each named path
|
|
@@ -55,9 +62,9 @@ the path and the change the audit reports. Run the audit again after each accept
|
|
|
55
62
|
If the audit still fails, or starts with `Status: blocked`, route out and stop:
|
|
56
63
|
|
|
57
64
|
```text
|
|
58
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
65
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
59
66
|
```
|
|
60
67
|
|
|
61
68
|
When the audit passes, run
|
|
62
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
69
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/complete` with the actual browser
|
|
63
70
|
outcome.
|
|
@@ -9,17 +9,18 @@ implementation, and proof to separate subagents within the target project.
|
|
|
9
9
|
Run this before anything else in this prompt:
|
|
10
10
|
|
|
11
11
|
```text
|
|
12
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
12
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard identify --coding-agent <coding-agent-slug> --model <model-id> --model-family <model-family>
|
|
13
13
|
```
|
|
14
14
|
|
|
15
15
|
Report your product slug, such as `codex` or `claude-code`. The command needs no sign-in,
|
|
16
16
|
browser, or project. If this run already ran `onboard identify`, skip this step.
|
|
17
17
|
|
|
18
|
-
- `--model`: copy the model id exactly as your instructions write it.
|
|
19
|
-
|
|
20
|
-
- `--model-family`:
|
|
21
|
-
|
|
22
|
-
|
|
18
|
+
- `--model`: copy the model id exactly as your instructions write it. Without an exact
|
|
19
|
+
model id, leave out `--model` rather than guess.
|
|
20
|
+
- `--model-family`: give it even without `--model`. Closest of: `claude-opus`,
|
|
21
|
+
`claude-sonnet`, `claude-haiku`, `gpt-5`, `gemini`, `grok`, `glm`, `minimax`, `mimo`,
|
|
22
|
+
`deepseek`, `qwen`, `kimi`, else the model's series. Leave it out if you do not know it.
|
|
23
|
+
- Never give your product slug as a model, or as an unlisted family.
|
|
23
24
|
|
|
24
25
|
## Waits, stops, and subagents
|
|
25
26
|
|
|
@@ -37,13 +38,13 @@ carry on. Keep a read-only assignment read-only. If no subagent works at all, re
|
|
|
37
38
|
once:
|
|
38
39
|
|
|
39
40
|
```text
|
|
40
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
41
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase delegation-unavailable
|
|
41
42
|
```
|
|
42
43
|
|
|
43
44
|
Before you end your turn to wait for the developer, report the pause:
|
|
44
45
|
|
|
45
46
|
```text
|
|
46
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
47
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
47
48
|
```
|
|
48
49
|
|
|
49
50
|
A turn that ends while the run waits is a pause, not a stop, even when your harness ends it.
|
|
@@ -53,7 +54,7 @@ If you stop this run, send one short report before you stop. The friction comman
|
|
|
53
54
|
the telemetry setting the developer already chose, so it needs no separate question.
|
|
54
55
|
|
|
55
56
|
```text
|
|
56
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
57
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard friction --phase stop --category <slug> --message "<sentences>"
|
|
57
58
|
```
|
|
58
59
|
|
|
59
60
|
`--message` takes one or two sentences: the step you stopped at and what stopped it.
|
|
@@ -67,7 +68,7 @@ one scoped fix, and sends the report itself when no fix is approved.
|
|
|
67
68
|
Before any subagent or project process runs, capture the developer's existing work:
|
|
68
69
|
|
|
69
70
|
```text
|
|
70
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
71
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect
|
|
71
72
|
```
|
|
72
73
|
|
|
73
74
|
Keep the exact protected path list it prints and give that list to every subagent. No
|
|
@@ -82,20 +83,20 @@ the current frontend-to-runtime-to-agent round trip and inspect `/info`.
|
|
|
82
83
|
Wait for the inspection subagent to finish. Then report that the inspection came back:
|
|
83
84
|
|
|
84
85
|
```text
|
|
85
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
86
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase research-returned
|
|
86
87
|
```
|
|
87
88
|
|
|
88
89
|
If the inspection did not prove the baseline this intent extends, stop here without
|
|
89
90
|
changing files:
|
|
90
91
|
|
|
91
92
|
```text
|
|
92
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
93
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
93
94
|
```
|
|
94
95
|
|
|
95
96
|
When the inspection proved what this intent needs, read the next prompt now, in this turn:
|
|
96
97
|
|
|
97
98
|
```text
|
|
98
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
99
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/rich-threads/plan
|
|
99
100
|
```
|
|
100
101
|
|
|
101
102
|
That prompt holds the plan, the developer's approval, and the paths the plan changes.
|
|
@@ -20,7 +20,7 @@ a `start` with no `--run` reprints the id this stopped run already holds, and on
|
|
|
20
20
|
covering both reads as one run that did two different things.
|
|
21
21
|
|
|
22
22
|
```text
|
|
23
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
23
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard start --run <new-12-character-id>
|
|
24
24
|
```
|
|
25
25
|
|
|
26
26
|
Do not read a generic onboarding node from this run. A feature run and a generic run are
|
|
@@ -45,10 +45,16 @@ Name the one fix you propose, and say which file, setting, or process it touches
|
|
|
45
45
|
developer whether they approve that one fix. If they decline, or the question cannot be
|
|
46
46
|
put to them, stop here and send the report below.
|
|
47
47
|
|
|
48
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
49
|
+
|
|
50
|
+
```text
|
|
51
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
52
|
+
```
|
|
53
|
+
|
|
48
54
|
If they approve it, make that one fix and nothing else. Then come back into this run:
|
|
49
55
|
|
|
50
56
|
```text
|
|
51
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
57
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard resume --message "<approval>"
|
|
52
58
|
```
|
|
53
59
|
|
|
54
60
|
Put the developer's approval in `--message`, in one or two sentences: the fix they
|
|
@@ -69,7 +75,7 @@ Send one short report. The friction command follows the telemetry setting the de
|
|
|
69
75
|
already chose, so it needs no separate question.
|
|
70
76
|
|
|
71
77
|
```text
|
|
72
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
78
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard friction --phase stop --category <slug> --message "<sentences>"
|
|
73
79
|
```
|
|
74
80
|
|
|
75
81
|
`--message` takes one or two sentences: the step you stopped at and what stopped it.
|
|
@@ -14,7 +14,7 @@ voice proof. Record changed paths and validation results.
|
|
|
14
14
|
Before you spawn the implementation subagent, take the authorized list from the CLI:
|
|
15
15
|
|
|
16
16
|
```text
|
|
17
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
17
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard audit
|
|
18
18
|
```
|
|
19
19
|
|
|
20
20
|
From its result, copy the paths under `Authorized to modify:`. A passed or failed audit with
|
|
@@ -29,11 +29,18 @@ protected path, it must not edit it. It returns a result that starts with `Statu
|
|
|
29
29
|
and names the file under Blockers.
|
|
30
30
|
|
|
31
31
|
That result is a question for the developer. The file has not changed yet, so consent can
|
|
32
|
-
still go on the record. Name the file and why the work needs it
|
|
33
|
-
|
|
32
|
+
still go on the record. Name the file and why the work needs it.
|
|
33
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
34
34
|
|
|
35
35
|
```text
|
|
36
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
36
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
37
|
+
```
|
|
38
|
+
|
|
39
|
+
Then end your turn and wait for the developer's answer. If they allow it, record their
|
|
40
|
+
answer before any edit:
|
|
41
|
+
|
|
42
|
+
```text
|
|
43
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"
|
|
37
44
|
```
|
|
38
45
|
|
|
39
46
|
Continue only when the result starts with `Status: passed`. Run the audit again, take the
|
|
@@ -45,7 +52,7 @@ feature stop route below.
|
|
|
45
52
|
After validation and each repair, run:
|
|
46
53
|
|
|
47
54
|
```text
|
|
48
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
55
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard audit
|
|
49
56
|
```
|
|
50
57
|
|
|
51
58
|
Continue only when it starts with `Status: passed`. A path under `Authorized to modify:` is
|
|
@@ -56,7 +63,7 @@ with the implementation subagent's `Files changed` section. If that section does
|
|
|
56
63
|
the path, accept the developer's external change:
|
|
57
64
|
|
|
58
65
|
```text
|
|
59
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
66
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --accept-external --path <path>
|
|
60
67
|
```
|
|
61
68
|
|
|
62
69
|
For an env file where the developer placed a requested credential, use
|
|
@@ -69,11 +76,11 @@ Run the audit again after each accepted change. If it still fails, or starts
|
|
|
69
76
|
with `Status: blocked`, route out and stop:
|
|
70
77
|
|
|
71
78
|
```text
|
|
72
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
79
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
73
80
|
```
|
|
74
81
|
|
|
75
82
|
If validation cannot pass, or this intent needs a prerequisite the app does not have, use
|
|
76
83
|
the feature stop route above without further changes.
|
|
77
84
|
|
|
78
85
|
Otherwise run
|
|
79
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
86
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/voice/proof`.
|
|
@@ -12,7 +12,7 @@ The approval is a wait for the developer. Before you end your turn to wait, repo
|
|
|
12
12
|
pause:
|
|
13
13
|
|
|
14
14
|
```text
|
|
15
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
15
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
16
16
|
```
|
|
17
17
|
|
|
18
18
|
A turn that ends while the run waits is a pause, not a stop, even when your harness ends it.
|
|
@@ -34,7 +34,7 @@ you plan. Run `onboard protect` again from the target app directory: on the same
|
|
|
34
34
|
prints the baseline the run already holds.
|
|
35
35
|
|
|
36
36
|
```text
|
|
37
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
37
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect
|
|
38
38
|
```
|
|
39
39
|
|
|
40
40
|
Then have one read-only subagent repeat the inspection for the findings above, and wait
|
|
@@ -50,25 +50,33 @@ the repository proves its variable name is configured. Otherwise, explain the se
|
|
|
50
50
|
take the feature stop route rather than writing a placeholder:
|
|
51
51
|
|
|
52
52
|
```text
|
|
53
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
53
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
54
54
|
```
|
|
55
55
|
|
|
56
56
|
Show a small plan for the documented transcription service, the existing composer, focused
|
|
57
57
|
tests, and a microphone/canned-transcript proof. Ask for approval as its own question.
|
|
58
58
|
|
|
59
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
60
|
+
|
|
61
|
+
```text
|
|
62
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-plan-approval
|
|
63
|
+
```
|
|
64
|
+
|
|
59
65
|
The plan must list every protected path it needs to change under `Authorization requested`,
|
|
60
|
-
with one sentence explaining why. Write `None` when it needs none.
|
|
61
|
-
|
|
62
|
-
|
|
66
|
+
with one sentence explaining why. Write `None` when it needs none. A step that installs or
|
|
67
|
+
upgrades a package changes its manifest, so list a protected manifest such as `package.json`
|
|
68
|
+
or `pyproject.toml`. Its consent covers its lockfile, beside it or at the workspace root.
|
|
69
|
+
Authorization covers changing a protected path, not removing it, so the plan must not
|
|
70
|
+
delete, move, or rename one.
|
|
63
71
|
|
|
64
72
|
After approval, record each approved path before implementation:
|
|
65
73
|
|
|
66
74
|
```text
|
|
67
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
75
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --authorize --path <path> --reason "<the plan's sentence>"
|
|
68
76
|
```
|
|
69
77
|
|
|
70
78
|
If an approved path changed after capture and no implementation step has run, add
|
|
71
79
|
`--with-prior-change`. Continue only when every result starts with `Status: passed`. Do not
|
|
72
80
|
authorize a path the approved plan did not list.
|
|
73
81
|
|
|
74
|
-
Then run `npx --prefer-offline --yes copilotkit@4.
|
|
82
|
+
Then run `npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/voice/implement`.
|
|
@@ -17,11 +17,18 @@ not edit it. It returns a result that starts with `Status: blocked` and names th
|
|
|
17
17
|
Blockers.
|
|
18
18
|
|
|
19
19
|
That result is a question for the developer. The file has not changed yet, so consent can
|
|
20
|
-
still go on the record. Name the file and why the fix needs it
|
|
21
|
-
|
|
20
|
+
still go on the record. Name the file and why the fix needs it.
|
|
21
|
+
Before you end your turn to wait for the developer, report the pause:
|
|
22
22
|
|
|
23
23
|
```text
|
|
24
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
24
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
25
|
+
```
|
|
26
|
+
|
|
27
|
+
Then end your turn and wait for the developer's answer. If they allow it, record their
|
|
28
|
+
answer before any edit:
|
|
29
|
+
|
|
30
|
+
```text
|
|
31
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect --authorize --unplanned --path <path> --reason "<what the developer said>"
|
|
25
32
|
```
|
|
26
33
|
|
|
27
34
|
Continue only when the result starts with `Status: passed`. Run the audit again, take the
|
|
@@ -34,11 +41,11 @@ Report the outcome of each attempt as it ends, before any repair. The command co
|
|
|
34
41
|
attempts and the repairs:
|
|
35
42
|
|
|
36
43
|
```text
|
|
37
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
44
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard proof --step round-trip --outcome <passed|failed|skipped>
|
|
38
45
|
```
|
|
39
46
|
|
|
40
47
|
Use `passed` only for proved voice input, `failed` for an attempted proof that failed, and
|
|
41
|
-
`skipped` when the proof could not run. Then run `npx --prefer-offline --yes copilotkit@4.
|
|
48
|
+
`skipped` when the proof could not run. Then run `npx --prefer-offline --yes copilotkit@4.25.0 onboard audit`.
|
|
42
49
|
Continue only when it starts with `Status: passed`.
|
|
43
50
|
|
|
44
51
|
If the audit fails, never repair, reset, or revert a protected path. Compare each named path
|
|
@@ -53,9 +60,9 @@ the path and the change the audit reports. Run the audit again after each accept
|
|
|
53
60
|
If the audit still fails, or starts with `Status: blocked`, route out and stop:
|
|
54
61
|
|
|
55
62
|
```text
|
|
56
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
63
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
57
64
|
```
|
|
58
65
|
|
|
59
66
|
When the audit passes, run
|
|
60
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
67
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/complete` with the actual surface
|
|
61
68
|
outcome.
|
|
@@ -8,17 +8,18 @@ read-only, implementation, and proof subagents, each restricted to the target pr
|
|
|
8
8
|
Run this before anything else in this prompt:
|
|
9
9
|
|
|
10
10
|
```text
|
|
11
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
11
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard identify --coding-agent <coding-agent-slug> --model <model-id> --model-family <model-family>
|
|
12
12
|
```
|
|
13
13
|
|
|
14
14
|
Report your product slug, such as `codex` or `claude-code`. The command needs no sign-in,
|
|
15
15
|
browser, or project. If this run already ran `onboard identify`, skip this step.
|
|
16
16
|
|
|
17
|
-
- `--model`: copy the model id exactly as your instructions write it.
|
|
18
|
-
|
|
19
|
-
- `--model-family`:
|
|
20
|
-
|
|
21
|
-
|
|
17
|
+
- `--model`: copy the model id exactly as your instructions write it. Without an exact
|
|
18
|
+
model id, leave out `--model` rather than guess.
|
|
19
|
+
- `--model-family`: give it even without `--model`. Closest of: `claude-opus`,
|
|
20
|
+
`claude-sonnet`, `claude-haiku`, `gpt-5`, `gemini`, `grok`, `glm`, `minimax`, `mimo`,
|
|
21
|
+
`deepseek`, `qwen`, `kimi`, else the model's series. Leave it out if you do not know it.
|
|
22
|
+
- Never give your product slug as a model, or as an unlisted family.
|
|
22
23
|
|
|
23
24
|
## Waits, stops, and subagents
|
|
24
25
|
|
|
@@ -36,13 +37,13 @@ carry on. Keep a read-only assignment read-only. If no subagent works at all, re
|
|
|
36
37
|
once:
|
|
37
38
|
|
|
38
39
|
```text
|
|
39
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
40
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase delegation-unavailable
|
|
40
41
|
```
|
|
41
42
|
|
|
42
43
|
Before you end your turn to wait for the developer, report the pause:
|
|
43
44
|
|
|
44
45
|
```text
|
|
45
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
46
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase awaiting-developer
|
|
46
47
|
```
|
|
47
48
|
|
|
48
49
|
A turn that ends while the run waits is a pause, not a stop, even when your harness ends it.
|
|
@@ -52,7 +53,7 @@ If you stop this run, send one short report before you stop. The friction comman
|
|
|
52
53
|
the telemetry setting the developer already chose, so it needs no separate question.
|
|
53
54
|
|
|
54
55
|
```text
|
|
55
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
56
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard friction --phase stop --category <slug> --message "<sentences>"
|
|
56
57
|
```
|
|
57
58
|
|
|
58
59
|
`--message` takes one or two sentences: the step you stopped at and what stopped it.
|
|
@@ -66,7 +67,7 @@ one scoped fix, and sends the report itself when no fix is approved.
|
|
|
66
67
|
Before any subagent or project process runs, capture the developer's existing work:
|
|
67
68
|
|
|
68
69
|
```text
|
|
69
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
70
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard protect
|
|
70
71
|
```
|
|
71
72
|
|
|
72
73
|
Keep the exact protected path list it prints and give that list to every subagent. No
|
|
@@ -76,26 +77,26 @@ approved authorization before implementation.
|
|
|
76
77
|
Before edits, inspect the runtime route, CopilotKit provider/chat composer, current agent,
|
|
77
78
|
existing provider credential variable names (presence only), package versions, browser
|
|
78
79
|
permissions approach, and test/dev commands. Prove the existing OSS baseline with `/info`,
|
|
79
|
-
`npx --prefer-offline --yes copilotkit@4.
|
|
80
|
+
`npx --prefer-offline --yes copilotkit@4.25.0 verify --expect-runtime oss --round-trip --agent <agent-id> --json`,
|
|
80
81
|
and a real UI request when a browser is available.
|
|
81
82
|
|
|
82
83
|
Wait for the inspection subagent to finish. Then report that the inspection came back:
|
|
83
84
|
|
|
84
85
|
```text
|
|
85
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
86
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard checkpoint --phase research-returned
|
|
86
87
|
```
|
|
87
88
|
|
|
88
89
|
If the inspection did not prove the baseline this intent extends, stop here without
|
|
89
90
|
changing files:
|
|
90
91
|
|
|
91
92
|
```text
|
|
92
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
93
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/stop
|
|
93
94
|
```
|
|
94
95
|
|
|
95
96
|
When the inspection proved what this intent needs, read the next prompt now, in this turn:
|
|
96
97
|
|
|
97
98
|
```text
|
|
98
|
-
npx --prefer-offline --yes copilotkit@4.
|
|
99
|
+
npx --prefer-offline --yes copilotkit@4.25.0 onboard read feature/voice/plan
|
|
99
100
|
```
|
|
100
101
|
|
|
101
102
|
That prompt holds the plan, the developer's approval, and the paths the plan changes.
|