The body is lean and assumes Claude's competence — it never explains what Docker, Nextflow, or Snakemake are — and every section is actionable. Version/date specifics are load-bearing authoritative pointers rather than padding, so most tokens earn their place; the inline time-sensitive details are a minor risk but do not push it to anchor 2.

Provides copy-paste-ready commands ('uv pip install "latch==2.76.8"', 'latch register --staging .', 'latch develop .') and a complete executable Python workflow example with imports and decorators, fully meeting the 'executable code/commands' anchor.

The 'Recommended Development Lifecycle' is a clearly numbered seven-step sequence with explicit validation checkpoints ('Re-run staging registration after changing the Dockerfile', 'Register deliberately', duplicate-registration exit status 2) and feedback guidance for error recovery.

A routing table maps nine needs to nine real, one-level-deep reference files (all verified present in references/), plus a real scripts/inspect_latch_sdk.py; the overview stays concise while details are clearly signaled and easy to navigate.

Lists multiple concrete actions ('Build, register, debug, and operate bioinformatics workflows') alongside the full Latch toolchain (Python SDK, CLI, Latch Data and Registry, Nextflow, Snakemake, MCP), matching the 'multiple specific concrete actions' anchor.

Explicitly answers both what (build/register/debug/operate workflows on Latch) and when via an explicit 'Use when...' clause, satisfying the highest anchor.

The 'Use when' clause covers natural user phrasings — 'authoring or deploying Latch workflows', 'configuring resources or interfaces', 'moving data', 'launching and monitoring runs' — giving good coverage of terms a user would actually say.

The bioinformatics-on-Latch niche with platform-specific triggers (Latch Data, Registry, Latch MCP) is distinct and unlikely to fire for unrelated skills.