live-e2e-verify
Skill0 starsChanged 27 days agoMITView on GitHub
Live verification pattern for {{PROJECT}} — use before declaring any phase/feature "done", especially anything touching the API, database, permissions, storage, or AI pipeline. Author pass runs against your local stack; the merger repeats it against the live one. Throwaway identities, full cleanup.
In this repository5 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- persona-reviewSkill
- session-contextSkill
- worklogSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
Write a script in the session scratchpad (never committed):
≈ “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.
Files
Live E2E Verification ({{PROJECT}})
Not "done" until seen working against a real stack. Typecheck passing and code looking right is not verification. Two passes, same script:
- Author pass — you, against your local stack (setup in
docs/ONBOARDING.md). No production credential needed. - Merger pass — whoever merges, against the live stack (needs whatever named, per-person, revocable credential the project hands out). Same script, same assertions, same cleanup. This is the only sanctioned write pattern against production from a laptop.
Standard sequence per phase/feature
{{TYPECHECK_CMD}}— clean across all workspaces.{{TEST_CMD}}— unit suite passes.- {{BUILD_CMD — the step that proves the client bundles/builds; skip if server-only}}.
- Live E2E script (below) for anything touching API/DB/storage/AI.
- Browser or device walk for anything with a UI: drive the real screen, assert the backend state after every write (not just the screen), keep screenshots as evidence.
Live E2E pattern
Write a script in the session scratchpad (never committed):
- Create a throwaway identity named
{{project}}-verify-<random>with the admin credential (local: from the local stack; live: your named key). - Sign in for a real token and exercise the endpoints as that identity — the same path production clients take.
- Assert all of:
- Status codes for happy AND sad paths (400 bad input, 401 no auth, 404 foreign/missing resource, 422/429/5xx where the contract says so).
- Rows actually written, with expected values (query with the admin credential).
- Permissions both directions: owner sees their rows; a second throwaway sees zero and gets 404 on cross-user access by id. Storage: prefix/ownership policy blocks reads and writes outside the identity's own prefix.
- Concurrency where it matters (N parallel spends at limit L → exactly L succeed).
- Delete the throwaway identities at the end — including anything cascades do NOT reach (storage objects, external-provider records, queued jobs). Never leave test data behind, even on script failure (trap/finally).
Gotchas that bite during verification
<!-- Promote here from worklogs the moment a trap bites twice. Keep each to 2–3 lines with
the date. Typical residents: client-SDK calls that behave differently stateless vs
stateful; model/provider access ≠ having a key; empty-string env vars treated as unset;
scheduled jobs that no-op locally and must be driven by hand; analytics events that
must be asserted against a local sink, never the real host. -->- {{gotcha}} ({{date}})
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
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.