agentleFS
randommoniclerandommonicle/claude-skills/one-real-ride/SKILL.md

one-real-ride

Skill27 starsChanged 19 days agoApache-2.0View on GitHub

Before declaring any deployed or user-facing change done, shipped, or closed out, run one real end-to-end invocation through the real seams — the deployed function, a real login and JWT, the real model call, the real browser or device. Green CI, green smokes, and a clean typecheck prove the layers; only the ride proves the seams. Triggers on "done", "shipped", "live", closing out a feature, and after any migration that existing code reads. Does not fire mid-implementation on individual command results (verify-the-effect's norm covers per-signal checks), nor on changes with no deployed or user-facing surface.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before "done": one real end-to-end invocation of the changed path, through the real seams — the deployed function (not the local one), a …

    line 15
  2. Very specific:

    A deploy record names the tree it gated: Before each deploy, check the tree against the gated range (git diff --name-only <base> HEAD …

    line 51

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

One real ride

Seven incidents where every automated gate was green and the feature was broken in production or the UI, plus a schema change that broke every report for five days while tsc and the whole test suite stayed green. The pattern is always the same: each layer was proven alone, and the defect lived in a seam between them. Layer: the done-gate hub.

The rule

Before "done": one real end-to-end invocation of the changed path, through the real seams — the deployed function (not the local one), a real authenticated session (not a service-role shortcut), the real model call, the real browser or the real device. The smoke proves the layer; the ride proves the seams.

What counts as a seam

  • Deployed runtime vs local: a deployed Edge Function 500'd on a dropped column CI could not see.
  • Auth: a function's own role pre-check was wrong in a way its smoke structurally could not catch.
  • Type-check vs render: "tsc green is not a working UI" — twice, including a Rules-of-Hooks blank page.
  • Build vs dev server: they run different code paths; a green production build does not prove the dev server boots, and vice versa.
  • Merge vs ship: know the deploy path per change type — a server merge may be live in minutes while an app merge reaches nobody until the next signed build. A merge is never a device test; the device test is the gate for device- or WebView-specific code.
  • Inherited guards and configs: anything borrowed from a sibling surface was calibrated for its original inputs — the first real invocation is what catches the mis-calibration three reviews missed.
  • Schema: after any migration existing code reads, ride the highest-value path once (generate the report, make the request). It is the only check that catches what synthetic-data tests cannot.

Routes

  • Interpreting any individual success signal during the ride → verify-the-effect (the norm; its playbook holds the diagnostics discipline).
  • The change sends transactional email → email-delivery-verification.
  • The change involved env vars, CI secrets, or hosted config → env-change-verification.
  • The change touched build config, lockfiles, or toolchain → reproduce-the-real-build.

What this skill does not do

It does not replace tests, smokes, or catalog verification — it is the last gate after them, not a substitute. It does not fire on library-internal changes with no deployed surface.

Why

Integration bugs live only in the seams, and every seam incident in the corpus was discovered by a user or a first real invocation — never by the green board. Evidence: PropOS LESSONS_LEARNED Sessions 13, 19, 23, 27, 43, 2026-07-05, 2026-07-11; ASH LESSONS_LEARNED 2026-05-27, 2026-07-11.

Additions from the misses-log sweep (2026-10-22)

  • A deploy record names the tree it gated. Before each deploy, check the tree against the gated range (git diff --name-only <base> HEAD prints only that wave's paths); two waves on one base deploy one at a time. (AI domain L-103.)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.