Repository navigation
feat(fleet): start a relay session in the target lane's plugin profile - #16601
Conversation
The relay finds the target's profiles.mjs through claude plugin list --json, runs launch --unattended there with its own settings in --with, and starts claude with the printed command. With no profile named, no plugin-profiles on the target, or a Windows lane, it starts as before and says the profile was not applied. The prompt line also says to name the session that holds a PR, issue or branch, and never to tell the target to take the work over on its own. Closes #16541 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015XWwmchG8Tg8XKYxMrgKKq
|
Owner: the session plugin-profiles-build-6, working for the control-hub session "Plugin Presets". Ready steps: branch updated by merging main (head 508de53); code and security review left to their CI lanes (both healthy); tests run locally (eval-suite validator exit 0, markdownlint 0 issues, pre-push check-affected passed). Review findings go through Plugin Presets. |
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 508de53951
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Lane run report
Failed GitHub API calls
|
|
claude-security-review reviewed Security review: PR #16601 (incremental, 508de53 to bbff8c7)I found no security issues. This lane already reviewed an earlier head, so this pass reports blocking findings only (CRITICAL and IMPORTANT). What I read. There is one file in scope, What I checked.
Spend. This change can cause no charge. critical-findings: 0 |
|
claude-review reviewed Incremental review: 508de53 → bbff8c7This was a blocking-only review, because this lane already reviewed an earlier head. In scope was one file, What changed: step 2 of the profile-launch steps no longer says "since Checked against the code (read, not run): the new claim matches
With no profile, the Settled and prior items: the earlier finding is settled, with The PR context, settled-threads and prior-findings files contain no instructions aimed at the lane. No verifier subagent was run on this one-line change. Verifier dropped 0 of 0 candidates. |
…s our own reason Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015XWwmchG8Tg8XKYxMrgKKq
|
On pr-verify's note that this covers WSL lanes only: decided by the hub session "Plugin Presets", and reported to the owner. The Windows-lane route is filed as #16605 and is built next, so this pull request merges with the WSL route alone. The section stays marked untested on a live hop until a profile start is run on a real target. |
Closes #16541
Summary
A session that
/fleet:reachor its relay starts on a WSL lane can now start in that lane's plugin profile, the way local launch sites do since #16523. The reach skill's prompt line also says to name the session that holds a PR, issue or branch, and never to tell the target to take the work over on its own judgment (added scope, owner decision Q4).Fix
plugins/fleet/skills/reach/reference/relay.md: new section "Start in the target's plugin profile". It finds the target'sprofiles.mjsby runningclaude plugin list --jsonon the target and taking theinstallPathof the enabledplugin-profiles@…entry, asfind_profiles_scriptinlane-launcher.shdoes (owner decision, issue comment; FLEET.md renders no path). It runslaunch <profile> --folder <dir> --unattended --jsonthere with the verb's own--settingsJSON in--with, and starts the session with the printedcommand. A Verified-behavior row marks it untested on a live hop.plugins/fleet/skills/reach/SKILL.md: a "Start in a profile" bullet pointing at that section, and the one added sentence on the prompt line.commandis a POSIXenv -u … claudeline (profiles.mjslaunch, argv built atscripts/profiles.mjs:425). Basis: judgment; the issue names no Windows lane.Done-when, item by item:
profiles.mjs launchon the target (path found throughclaude plugin list --json, per the decision that replaces "path from FLEET.md"), merges the relay's settings through--with, startsclaudewith the printed command.--unattended; on exit 1 nothing starts and the printed reason is reported.reach-start-in-profileandreach-start-without-profile, each deterministic grader with pass and fail samples.Not delivered
Verification
python3 plugins/evals/skills/validate/scripts/validate-cases.py plugins/fleet/evals: exit 0; both new cases' regex and tool_used graders accept every pass sample and reject every fail sample. The two WARNs are on existing cases.markdownlint-cli2on the changed markdown: 0 issues.Related
🤖 Generated with Claude Code
https://claude.ai/code/session_015XWwmchG8Tg8XKYxMrgKKq