Remove dropdown chevron icon from Svelte git branch component
Takeaway: I'd specify next time whether the branch icon and spinner should stay, since that ambiguity took a few extra verification steps.
Takeaway: I'd specify next time whether the branch icon and spinner should stay, since that ambiguity took a few extra verification steps.
Takeaway: Next time I'd be more explicit upfront that 'unfold to the same layer' meant equal treatment of tool and thinking blocks in their original order, not reordering tool calls ahead of thinking — that ambiguity cost a whole wrong implementation.
Takeaway: Next time I'd include a screenshot of the desired banner layout up front, since the first pass removed the right text but left an alignment issue I had to flag separately.
Takeaway: Next time I'd describe the icon inconsistency more precisely upfront — mentioning that both the open and closed folder states looked different sizes would have saved a round trip.
Takeaway: Next time I'd ask for the migration to be done component-by-component with a build check after each one, instead of generating the whole 25-component app in one shot and debugging the pile of runtime errors afterward.
Takeaway: Next time I'd start by asking directly for the two known root causes at once: the frontend title priority order and the 64KB session-summary read truncation on large image messages, since both turned out to matter.
Takeaway: Next time I'd ask the agent up front to inspect how @pierre/diffs actually themes its shadow DOM (CSS variables vs.
Takeaway: Next time I'd describe the bug with more specifics up front, like which sessions or scroll actions trigger it, since that would have saved a lot of exploratory code reading.
Takeaway: Next time I'd just ask directly for a container-query based responsive fix instead of describing the symptom, since that's what the agent eventually landed on anyway.
Takeaway: This was a one-line constant change, so I could have just asked for the exact constant name if I'd known it, but letting the agent grep for it first was fine too.
Takeaway: Next time I'd tell it upfront to keep some layering between the composer, the tool chips, and the send button instead of flattening everything to base00 in one pass — going all-in on base00 made the light theme look muddy.
Takeaway: Next time I'd upfront ask for the facts/vector split explicitly instead of letting the agent build the monolithic GraphModality first and then retrofitting — that second prompt ('还有 factsworker和vectorworker没实现') is what actually forced the right architecture.
Takeaway: Next time I'd point directly at toolBlock.ts and ChatTranscript.vue upfront since I already knew roughly where tool rendering logic lived, instead of letting the agent grep broadly first.
Takeaway: Next time I'd point directly at the CSS class name if I know it, since most of the time here went into searching the codebase for where the spacing was defined.
Takeaway: I would have started by pointing at the exact CSS classes and values I wanted instead of iterating turn by turn (10px, then 4px, then 0px) — that wasted a couple of build cycles.
Takeaway: I'd start by pointing the agent straight at useBridgeClient.ts and ws-rpc-adapter.ts instead of letting it discover the RPC layer through several rounds of grepping from the Vue component outward.
Takeaway: Next time I'd start by asking directly for a headless mode in /web instead of first exploring RPC/SDK paths that turned out to be dead ends — that exploration cost a lot of turns before we landed on the right approach.
Takeaway: Next time I'd specify upfront that packages should use package-name imports (not relative paths) and that internal packages must stay private, since that's exactly what I ended up correcting twice.