Sharpen /scalable grounding: per-claim, not per-session; do the cheap check yourself#13
Merged
Merged
Conversation
… yourself An agent running /scalable read every worker *around* the dashboard (app-router, website-supervisor, tunnel-auth, cloudflare.ts) and then ruled by transferring that architecture onto apps/web by analogy — never opening the one file the change actually routes through. The volume of neighboring grounding made the single load-bearing, unverified claim feel as solid as the rest. It then ended its turn asking the human to confirm what apps/web/wrangler.jsonc answers in seconds. The skill already said "read the actual code … before ruling" — the gap was framing, not a missing rule. Sharpen the "Ground it first" block: - Reframe "priors" to include conclusions minted this session from neighboring code — more dangerous than stale training data because they feel freshly earned. - New bullet: ground per claim, not per session — isolate the single load-bearing claim and verify that one directly; "built like its neighbors" is not evidence about the thing it touches. - New bullet: do the cheapest decisive check yourself — if one file settles it, open it; don't defer to the human. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
A coding agent running
/scalableon a dashboard routing change read every worker aroundapps/web— app-router, website-supervisor, tunnel-auth,cloudflare.ts— then ruled by transferring that architecture onto the dashboard by analogy, never opening the one file the change actually routes through (apps/web/wrangler.jsonc). Its own postmortem:The skill already said "Read the actual code and architecture this touches before ruling." The gap was framing, not a missing rule — so this sharpens the existing "Ground it first" block rather than adding a new section.
What changed (
commands/scalable.mdonly)No generated/mirror files touched —
build-rules.jsfans outCLAUDE.md, notcommands/scalable.md;node scripts/build-rules.js --checkpasses.🤖 Generated with Claude Code