tdk-starlark-perf

Installation
SKILL.md

TDK Starlark performance

Use this skill to diagnose and improve slow TDK/Starlark generation and Tilt startup.

Workflow

  1. Run tdk doctor and flisk doctor to establish a baseline.
  2. Time flisk up or tilt up and note which phase is slow (discovery, generation, build).
  3. Reduce work: start only needed stacks with flisk up <stack> or tilt up --focus.
  4. Avoid broad file scans in Starlark: narrow discovery.paths, exclude large directories, and keep service manifests minimal.
  5. Reuse cached golden images; do not force rebuilds with TDK_GOLDEN_REBUILD unless necessary.
  6. Keep .autogenerated/ outputs out of manual edits; regenerate via flisk config regenerate instead of patching.
  7. Prefer smaller service graphs: split large monoliths into stacks and avoid unnecessary dependsOn chains.

Guardrails

  • Do not hand-edit generated .autogenerated/ files.
  • Do not disable health checks or validation to gain speed.
  • Measure before and after; report observed timings.
Installs
1
GitHub Stars
2
First Seen
Oct 21, 2026
Security Audits