flutter-read-logs: add plugin for reading run logs as context

[papeye]

What

Adds a new flutter-read-logs plugin: a /read-logs skill that pulls the running app's latest flutter run output into context (auth flow, a crash, a race, "why did X happen at runtime"), plus a flutter-read-logs-usage routing skill and a README documenting setup.

Why

There was no way for Claude to see a running app's output — diagnosing runtime behaviour meant copy-pasting terminal logs. This captures flutter run output per-project and surfaces the relevant slice on demand. Piloted in a LeanCode app (meg-app) before extracting it here.

How it works

• Logs captured per-project to /tmp/flutter-<repo>.log, derived from --git-common-dir so the main checkout and all worktrees share one file. Overwritten each launch; lives in /tmp.
• Two capture formats, auto-detected: Zed script transcript and VS Code/Cursor dart.dapLogFile (DAP; app output extracted from event:"output" lines via a grep-prefiltered python/jq pass).
• Reading is task-led, not a flat tail: reads the whole short run, or investigates a long one against the prompt (errors, event ordering, a feature…). Errors are one lens, not the only one.
• Flags stale runs (age + Dart files changed since capture) and, on first use, guides editor setup (and offers to apply it) — VS Code/Cursor dapLogFile keeps F5; Zed wraps the run task in script (macOS + Linux forms). Never commits; warns before editing a tracked config file. Sensitive-data caution included.

Notes

• Setup lives canonically in the plugin README.md; the read-logs skill repeats it inline (with a note) because it applies setup at runtime — the sanctioned exception in the contributor guide.
• Draft until I run a VS Code end-to-end pass (auto-add dapLogFile → simulator F5 → read) and CI is green. Couldn't run go run ./cmd/validate-plugins locally (no Go); relying on CI.

Ad

i_created_a_claude_code_comamn.mp4

[Copilot]

Pull request overview

Adds a new flutter-read-logs plugin to the LeanCode Claude Code plugins set, enabling on-demand ingestion of the most recent flutter run output (captured per-project under /tmp) so runtime behavior can be investigated without manual log copy/paste.

Changes:

• Introduces the /read-logs skill plus a flutter-read-logs-usage routing/help skill.
• Adds plugin documentation (plugins/flutter-read-logs/README.md) and initial changelog/manifest.
• Registers the plugin in the root README and marketplace manifest.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
README.md	Adds flutter-read-logs to the top-level plugin list.
plugins/flutter-read-logs/skills/read-logs/SKILL.md	New /read-logs skill with path resolution, format normalization, and setup guidance.
plugins/flutter-read-logs/skills/flutter-read-logs-usage/SKILL.md	New usage/routing skill for setup and when-to-use guidance.
plugins/flutter-read-logs/README.md	Plugin-level README describing setup and workflow.
plugins/flutter-read-logs/CHANGELOG.md	Initial plugin changelog entry.
plugins/flutter-read-logs/.claude-plugin/plugin.json	Plugin manifest for Claude Code marketplace tooling.
.claude-plugin/marketplace.json	Registers the new plugin and its skills in the marketplace index.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[wzslr321]

Me after seeing the AD

[wzslr321]

Structure looks good. It will be necessary to share it on Slack, so we can have more people testing it on different IDEs via Claude Code plugins.

I like the idea of having it lightweight via skill, while Marionette is still referenced as more proactive tool.

One open question to consider, but non-blocking and we probably will need to address it from marionette perspective as well.