Repository navigation
feat: kitbash import — reverse-compile a repo's existing agent configs (0.16.0) - #100
Merged
Merged
Conversation
…s (0.16.0)
Grounded in a landscape scan (113-agent research sweep): the strongest, most-
cited pain for teams running several coding agents is config drift and painful
onboarding across agents — named verbatim by 5+ independently-built sync tools.
Kitbash made you author a fresh skill; now it starts from what you already have.
- kitbash import: reads a repo's existing agent instruction files (CLAUDE.md,
AGENTS.md, GEMINI.md, CONVENTIONS.md, .cursorrules, .cursor/rules/*.mdc,
.github/copilot-instructions.md, .clinerules, .windsurf/.devin/rules/*),
measures what each costs, DETECTS DRIFT — where the copies silently diverged —
and synthesizes one KSF skill from the version most agents agree on. --write
saves + pins it so 'kitbash compile' regenerates every target from one source
and ends the drift. --name sets the skill name. Non-destructive; skips purely
kitbash-generated files. New importers.ts module (format readers + drift
grouping); cmdImport orchestrates + synthesizes; wired as a real command.
- honesty fix: corrected an overstated benchmark-page claim ('the number nobody
measures' — other tools do estimate context-file cost). Kitbash's distinct
claim is measuring it per-target, from one source, and enforcing it as a
declared budget at compile time; the page now says that.
337 assertions (+14): empty repo, agree/drift detection, --write + pin, the
imported skill compiles, dup-name refusal, kitbash-generated files skipped.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
github-actions
Bot
added
documentation
This branch was successfully deployed
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
1 participant
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 this, why now
A 113-agent research sweep of what multi-agent teams actually need found the strongest, most-cited pain is config drift and painful onboarding across agents — named verbatim by 5+ independently-built sync tools (amtiYo/agents, rule-porter, RuleSync, …) and a JetBrains survey (~5.3 hrs/month lost reconciling mismatched agent instructions). Kitbash made you author a fresh skill — the worst possible ask for a repo that already has the mess. This is the on-ramp: import what you already have.
kitbash importReverse-compile. Reads a repo's existing agent instruction files — CLAUDE.md, AGENTS.md, GEMINI.md, CONVENTIONS.md,
.cursorrules,.cursor/rules/*.mdc,.github/copilot-instructions.md,.clinerules,.windsurf/.devin/rules/*— and:--writesaves + pins it, sokitbash compileregenerates every target from that one source and ends the drift.--namesets the name. Non-destructive (leaves your originals in place); skips purely kitbash-generated files so it never re-imports its own output.New
importers.ts(format readers + drift grouping),cmdImportorchestrates, wired as a real command (not a stub).Honesty fix
Corrected an overstated benchmark-page claim — "the number nobody measures." Other tools (ctxlint) do estimate context-file token cost. Kitbash's genuinely distinct claim is measuring it per target, from one source, and enforcing it as a declared budget at compile time — the page now says that. (Same class as the credibility batch; the research surfaced it, so it gets fixed.)
Tests
337 assertions (+14): empty repo, agree/drift detection, file grouping,
--write+ lockfile pin, the imported skill actually compiles (loop closed), duplicate-name refusal, and purely-kitbash-generated files correctly skipped.