A 26K-star markdown skill library that bolts senior-engineer SDLC scaffolding (spec, plan, test, review, ship) onto AI coding agents like Claude Code.
Key Takeaways
Skills are executable workflows with checkpoints and exit criteria, not prose essays; that distinction is what makes them agent-actionable.
Anti-rationalization tables pre-write rebuttals to excuses agents (or tired engineers) use to skip steps like writing a spec or failing test first.
Six lifecycle phases map to slash commands: /spec, /plan, /build, /test, /review, /ship, plus /code-simplify.
Progressive disclosure via a router meta-skill (using-agent-skills) activates only relevant skills per phase, avoiding context poisoning.
Skills encode Google SWE practices: Hyrum’s Law, DAMP tests, ~100-line PR sizing, trunk-based dev, Chesterton’s Fence, and shift-left CI.
Hacker News Comment Review
Early adopters are selectively installing specific skills (api-design, ui-testing) rather than the full set, suggesting modular value over all-or-nothing adoption.
Concern about plugin removal difficulty signals that vendoring raw markdown into .cursor/rules/ or an AGENTS.md may be the more practical path for many teams.
Notable Comments
@y-curious: Plans to steal patterns but skipping the plugin install due to deletion risk; prefers customizing individual skills per developer.
