@outerlayer/cli 0.39.0-next.20261113235543 → 0.39.1-next.20261114003903
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/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,32 @@
|
|
|
1
1
|
# @outerlayer/cli
|
|
2
2
|
|
|
3
|
+
## 0.39.0
|
|
4
|
+
|
|
5
|
+
### Minor Changes
|
|
6
|
+
|
|
7
|
+
- 0032eac: A build that continues the session of a stopped build starts with that build's files. When a job ends before it ran to its end, the runner keeps the edits it did not commit and the commits it did not push, with the agent session on the runner's host. A job that continues the session gets them back, also in a new sandbox or after the kept disk is gone.
|
|
8
|
+
- cbc6e2e: Add `outerlayer work mock list`, `add` and `replace`.
|
|
9
|
+
|
|
10
|
+
### Patch Changes
|
|
11
|
+
|
|
12
|
+
- c809092: The code review reports only defects that the change adds or causes. It leaves out a defect that the code had before the change, and a problem in unchanged code names the changed line that causes it.
|
|
13
|
+
- 6eeb985: The reason of an incomplete build names the checks that do not pass before the cause of a proof recording that failed. The cause names its criterion. A long cause is cut, so the 200-character limit drops no check.
|
|
14
|
+
- 576065d: The criteria judge reads the repository a build pushed in. A build that changes only a secondary repository of its claim is judged, instead of failing with "the build pushed no branch".
|
|
15
|
+
- e35b658: A build that continues a session gets back the work of a build whose agent ended with no pull request. The runner kept that work only for a build that was stopped, failed or reached its time limit.
|
|
16
|
+
- 5174b82: A sandbox build gets a disk no smaller than its snapshot's. The runner reads the snapshot's disk from the sandbox provider, so a build whose `runner.build.disk` or factory machine has a smaller disk still starts.
|
|
17
|
+
- 60adc8c: The runner no longer reads a build's command while it is still starting it. A command that ended before its start call returned ended the build with the stage `provision`, and a slow start could end a build as interrupted.
|
|
18
|
+
- 1178c5a: The runner asks GitHub again when it refuses a token that the gateway issued a moment before. A build no longer ends at provision with `cannot read <repository>` for a refusal that passes seconds later.
|
|
19
|
+
- d0bd8de: `outerlayer run topics` prints the topics that a schedule automation already filed issues for, one on each line, so the agent of a run can use a topic's exact text.
|
|
20
|
+
- 6a841c0: A runner removes the lock files that a stopped git process left in its own repository cache and control-plane cache before it fetches. A build no longer fails its provision with `repository_unreadable` or `context_unavailable` because of a leftover `shallow.lock` after the runner stops during a fetch.
|
|
21
|
+
- 2b88ad9: A runner can build in sandboxes that OuterLayer provides: set `runner.sandbox.provider` to `"outerlayer"` and choose `runner.sandbox.machine`. The runner needs no sandbox provider account, its credentials stay on its machine, and the build's time counts as managed build-hours.
|
|
22
|
+
- 6e81b60: The runner reports protocol 56: it accepts the `review` claim kind in a gateway's answers and asks for no review work.
|
|
23
|
+
- bbc9655: A runner tells OuterLayer in its heartbeat whether it takes the runs of schedule automations, so OuterLayer can say why a run waits. A runner takes none when it is a managed or fleet runner, when OuterLayer provides its sandboxes, when its hooks are its own, when it has no command for a run, or when `runner.queues` lacks `implement`.
|
|
24
|
+
- 24ea1cb: The runner takes code review jobs: it reviews each requested commit of a pull request in a new container or microVM, with no setup steps, and reports problems with no pass or fail. The review's instructions move to the new maintained skill `outerlayer-review`, and the judge stage runs no code review.
|
|
25
|
+
- 4cad1a1: A sandbox build that uses the sandbox's Docker daemon stays on the sandbox's own network. It reaches the ports that its containers publish, such as a local Supabase database, on `127.0.0.1`. The runner's log says that the build runs no network helper, and why.
|
|
26
|
+
- 4ab9327: A running runner does not use a config file that changes `runner.sandbox.provider`. It keeps the last good settings and logs that a restart is needed. Before, it used the new settings with the client of the old provider, and each sandbox build failed at provision.
|
|
27
|
+
- 3650afa: The build and evidence skills name the item page's Criteria tab, which was the Review tab.
|
|
28
|
+
- 4cdc8ff: The runner waits up to 20 minutes for a sandbox that the sandbox provider is slow to start, then builds in it. A slow start ended the build at provision after 30 seconds.
|
|
29
|
+
|
|
3
30
|
## 0.38.0
|
|
4
31
|
|
|
5
32
|
### Minor Changes
|
package/dist/build-info.json
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"id":"2026-11-14T00:
|
|
1
|
+
{"id":"2026-11-14T00:46:07.942Z","root":"/home/runner/work/outerlayer/outerlayer","pkg":"outerlayer"}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@outerlayer/cli",
|
|
3
|
-
"version": "0.39.
|
|
3
|
+
"version": "0.39.1-next.20261114003903",
|
|
4
4
|
"description": "OuterLayer CLI — commands: init, claude, opencode, connect, doctor, login, logout, sync, daemon (alias: watch), emit, demo, findings, work, run, spec, runner, hooks, mcp, context. Capture and sync coding-agent sessions, and record proof of factory work.",
|
|
5
5
|
"license": "Apache-2.0",
|
|
6
6
|
"author": "Magu Studios, Inc.",
|
|
@@ -78,13 +78,13 @@ repository. Commit and push in each checkout you change, to its own branch only.
|
|
|
78
78
|
issue. It is the spec: find each thread's criterion there. Read the pull
|
|
79
79
|
request and its evidence comment. Read every person's
|
|
80
80
|
comment since the last agent comment, literally.
|
|
81
|
-
With an earlier run in your conversation, read only what changed
|
|
82
|
-
([references/earlier-run.md](references/earlier-run.md)).
|
|
81
|
+
With an earlier run in your conversation, read only what changed and
|
|
82
|
+
retest its failures ([references/earlier-run.md](references/earlier-run.md)).
|
|
83
83
|
Then read `checks` in the status JSON. A failing check lists its check
|
|
84
84
|
items under `items`, each with a `result` and a `reason`. When a person asks
|
|
85
85
|
you to fix failing checks or criteria, fix each failing check item by its reason.
|
|
86
86
|
4. **Decide each thread.** Reply on every criterion thread waiting on an
|
|
87
|
-
agent, with `--criterion`. Reply on every source thread with `--thread`.
|
|
87
|
+
agent, with `--criterion`. Reply on every source thread with `--thread`. Pick one per note:
|
|
88
88
|
- **Answer it.** A question gets an answer. Change nothing. Name the
|
|
89
89
|
file or test you read.
|
|
90
90
|
- **Fix it.** A requested change gets a fix. Change the code, tests, docs or
|
|
@@ -132,9 +132,8 @@ repository. Commit and push in each checkout you change, to its own branch only.
|
|
|
132
132
|
sentence of fact, then the decision a person must make. For a spec problem,
|
|
133
133
|
quote the criterion and give the replacement wording. `--ready` hands the
|
|
134
134
|
thread back to a person.
|
|
135
|
-
9. **Check nothing is left waiting.** Reread threads.
|
|
136
|
-
on an agent
|
|
137
|
-
or report it. Run `outerlayer sync` and `outerlayer work proofs`. Fix every
|
|
135
|
+
9. **Check nothing is left waiting.** Reread threads. Fix or report a thread
|
|
136
|
+
still waiting on an agent: its reply did not land or was flagged `unchanged`. Run `outerlayer sync` and `outerlayer work proofs`. Fix every
|
|
138
137
|
listed row that is not current.
|
|
139
138
|
10. **Report.** Write one line per thread: the criterion (or the general
|
|
140
139
|
thread), and fixed (with the replacement) or handed off (with the reason).
|
|
@@ -10,6 +10,12 @@ Do not read them again. Read only what changed since your last comment:
|
|
|
10
10
|
- new commits on the pull request's branch;
|
|
11
11
|
- new check results.
|
|
12
12
|
|
|
13
|
+
What the earlier run found the build could not do is out of date. You run in
|
|
14
|
+
a new build, and its runner version or host can differ. Examples are "the app
|
|
15
|
+
does not start", "a port is not reachable" and "the browser does not run".
|
|
16
|
+
Before you act on such a failure, or repeat it in a comment, run its command
|
|
17
|
+
again in this build. Report it only when it fails again, and quote the output.
|
|
18
|
+
|
|
13
19
|
Paths from the earlier run can be gone. One example is the governed context's
|
|
14
20
|
directory, which names a commit. Use the current paths and the current skills.
|
|
15
21
|
Do not use a path or a skill text from the earlier run.
|
|
@@ -144,6 +144,12 @@ its `instructions` say what the member wants. Keep the state file, re-orient
|
|
|
144
144
|
as rule 4 says, and go on from the phase the file names. A rule of the team's
|
|
145
145
|
`build` skill that stops a session which already ran a build does not apply.
|
|
146
146
|
|
|
147
|
+
What the earlier build found it could not do is out of date. This build is
|
|
148
|
+
new, and its runner version or host can differ. Examples are "the app does not
|
|
149
|
+
start", "a port is not reachable" and "the browser does not run". Before you
|
|
150
|
+
act on such a failure, or repeat it in a comment or a report, run its command
|
|
151
|
+
again in this build. Report it only when it fails again, and quote the output.
|
|
152
|
+
|
|
147
153
|
A session that continues also has `OUTERLAYER_WORK_RESTORED` when the earlier
|
|
148
154
|
job left edits it did not commit or commits it did not push. `1` means each
|
|
149
155
|
checkout is as that job left it: the same commit, with the edits not
|