scott-garvin
UserReusable Claude Code skills that let AI agents work safely on production systems.
Categories
Indexed Skills (6)
verify-before-claim
Use before reporting that any change, write, or fix succeeded. Confirm the result from an independent read rather than trusting the operation's own return value, know the specific ways a "success" lies (read-after-write lag, 200-but-not-persisted, wrote-a-different-field, 400-but-it-landed, permission-blocked verb that fails silently, green-by-skip test runs), and when the result genuinely cannot be verified, report "cannot determine, here is why" instead of a confident guess. Triggers on verifying a write, "did this work", an API or database mutation, "the change is done", or reporting a fix as complete.
build-test-truth
Use when building a project, running its tests, or reasoning about a CI result. A passing command is not a passing suite. Knows the "green-by-skip" traps - a run that discovered or executed zero tests, a suite that self-skipped on a missing environment flag, a tool/flag combination that reports success while running nothing, or the wrong test project running - and applies an execution gate that asserts the tests actually covering the change ran and passed. Triggers on build, compile, run the tests, CI, "the tests pass", vstest/xunit/jest/pytest, green build.
knowledge-registry
Use when an investigation is wasted by a misread data model, a confusing legacy behavior, or a non-obvious gotcha - and when writing code near one. Maintains a living registry of the traps that have burned past investigations (easy-to-misread models, misleading names, load-bearing quirks) so the same mistake never costs time twice. Read the relevant entry before diving into a known-treacherous area; add an entry whenever a surprise costs real time. Triggers on "this burned me before", confusing or legacy behavior, a field/status that doesn't mean what it says, repeated investigations into the same area.
prod-error-triage
Use to find and triage a production error or incident. Starts from the authoritative error source - the channel that actually captures every unhandled exception - rather than the dashboard everyone assumes is complete but is silently empty. Covers locating the real source, mapping an error to the environment and tier that produced it, filtering known baseline noise, and separating a brand-new incident from recurring background failures before reading the stack. Triggers on prod error, production exception, "why did X fail in prod", incident, stack trace triage.
release-cutover
Use to run a production release or cutover. Treats a release as an ordered checklist where every step is reversible and gated - apply schema migrations through an error-halting pipeline, smoke-test on a staging target before it takes traffic, swap with the platform's mandatory restart, promote background workers in lockstep, and keep a rehearsed rollback for every step. Composes the safe-migrations and production-write-gate skills. Triggers on "cut the release", "ship to prod", deploy, slot swap, cutover, go-live, rollback.
safe-migrations
Use when changing a database schema or seed/lookup data - adding or altering tables, columns, indexes, constraints, stored procedures, functions, or one-time data fixes. Every change is authored as an idempotent forward migration with a matching reverse (rollback) migration, keeps each object's canonical source file authoritative, and is applied through the migration pipeline behind the production-write-gate, never by editing production directly. Triggers on schema change, migration, DDL, add/alter column or table or index, stored procedure, seed data, or one-time data fix.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.