@outerlayer/cli 0.8.6 → 0.9.0

package/skill-pack/maintained/amend/SKILL.md

@@ -10,8 +10,10 @@ description: >
 
 Answer every thread waiting on an agent on one work item's pull request, then
 stop. A work item has one general thread and one thread per acceptance
-criterion. A person's fail, or a plain comment from a person, leaves a thread
-waiting on an agent. The person's newest comment says what they want. This
+criterion. A person's fail on the general thread leaves it waiting on an agent.
+A note a reviewer sent with Request changes (a sent note) leaves its criterion
+thread waiting on an agent. A plain comment from a person does not hand a
+thread to an agent. The person's newest comment says what they want. This
 skill never records a verdict. Only a person passes or fails a thread. A
 session can only reply.
 
@@ -43,25 +45,33 @@ straight to the git host. A local session uses the person's own git remote and
    `criterion` (null for the general thread), `waitingOn` (`agent`, `person` or
    null), `verdict`, `proof` and `comments`. Each comment carries `author`
    (`kind` is `person` or `agent`), `body`, `action` and `artifact`. Keep the
-   threads whose `waitingOn` is `agent`. If none is, say so and stop: there is
-   nothing to answer. Then read `outerlayer work status --item "$OUTERLAYER_WORK" --json`
-   for the pull request in `pullRequests` (`repository`, `prNumber`).
+   threads whose `waitingOn` is `agent`. One send of notes starts one run, so
+   expect several criterion threads and the general thread together. If none is
+   waiting, say so and stop: there is nothing to answer. Then read
+   `outerlayer work status --item "$OUTERLAYER_WORK" --json` for the pull
+   request in `pullRequests` (`repository`, `prNumber`).
 2. **Re-orient.** Read the issue the item names and the pull request, including
    its evidence comment, through the team's tracker tool or the git host. The
    issue body is the spec. Find each thread's criterion there. Read every
    person's comment since the last agent comment on the thread. Read them
    literally: they are the reviewer's instruction.
-3. **Decide each thread.** For each waiting thread, pick one:
-   - **Fix it.** The comment names something the change got wrong or did not
-     show. Change the code, tests, docs or evidence so the criterion is met as
-     written.
+3. **Decide each thread.** Reply on every criterion thread waiting on an
+   agent, with `--criterion`, so each note gets its answer under it. Read each
+   note and pick one:
+   - **Answer it.** A question gets an answer. Reply with the answer and
+     change nothing. Say which file or test you looked at.
+   - **Fix it.** A requested change gets a fix. The comment names something the
+     change got wrong or did not show. Change the code, tests, docs or evidence
+     so the criterion is met as written, and attach new proof when the
+     criterion declares one.
    - **The spec is wrong.** The comment shows the criterion cannot be met as
      written, for example because it names a path a repository rule forbids. A
      session never rewrites a criterion on its own. Hand it off (step 5), naming
      the exact wording change the owner should make.
-   - **Cannot tell.** The comment is ambiguous. Hand it off, quoting it and
-     saying what you would need to know.
-4. **Fix, prove, reply.** For each thread you fix:
+   - **Cannot tell.** The comment is ambiguous. Hand it off with `--ready`:
+     quote it and ask one question back.
+4. **Fix, prove, reply.** An answer needs no commit: reply on its criterion
+   thread with the answer and `--ready`. For each thread you fix:
    - Make the change on the pull request's branch. Keep it to what the comment
      asks. Do not refactor around it.
    - Run the tests that cover the criterion, then the repository's full

package/skill-pack/template/build/SKILL.md

@@ -194,5 +194,16 @@ Run them in order.
    Watch CI until every check has concluded.
 6. **Manual test and evidence.** Read [references/evidence.md](references/evidence.md)
    when the pull request opens, and run it. Capture the proof each criterion
-   declares, emit it, then end the run with the final report. The report
-   lists every candidate issue and every assumption.
+   declares, emit it. Then release the lease.
+7. **Release the lease.** Run this last, after the pull request is open and
+   the evidence is emitted, and run no other tool call after it:
+
+       outerlayer work release --item "$OUTERLAYER_WORK"
+
+   A session that names a work item holds a lease on it, and review stays
+   closed until the lease is released. In a session this command releases the
+   session's own lease and no one else's. A later tool call takes the lease
+   again, so write the final report as plain text once this has run. Skip this
+   step in a runner build, whose host releases its own lease. Then end the run
+   with the final report. The report lists every candidate issue and every
+   assumption.

package/skill-pack/template/build/references/evidence.md

@@ -2,8 +2,8 @@
 
 Read this when the pull request opens. The evidence agent starts then, in the
 same wave as the CI watch. Everything after it in this file runs once CI is
-conclusive and the evidence agent has returned. This is the last phase, and it
-ends the run: it deletes the state file and the report directory. The evidence
+conclusive and the evidence agent has returned. It deletes the state file and
+the report directory, then phase 7 releases the lease and ends the run. The evidence
 brief is in [agent-briefs.md](agent-briefs.md). Read every value from an
 earlier phase from the state file.