From 6f31422cecb9adf07f9b98dcdeeb3ac86efb25e5 Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Fri, 30 Oct 2026 08:34:55 -0700 Subject: [PATCH] docs(agents): never paper over a partial failure Co-Authored-By: Claude Opus 5.5 --- AGENTS.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/AGENTS.md b/AGENTS.md index 6271abeeec6..62b87bd63aa 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -15,7 +15,7 @@ Use this checkout's source and docs for repository work; published API reference - Pull the latest origin changes before working. - Dig into the root cause and fix it at the source. Never work around a fixable bug with a retry, sleep, or timeout, nor pass a flaky test by loosening its assertion or disabling production behavior. -- Fail loud and early. Error on unsupported or malformed input rather than warn and continue: supported or refused. +- Fail loud and early. Error on unsupported or malformed input rather than warn and continue: supported or refused. Never paper over a partial failure: when any part of an input or operation fails, fail the whole instead of skipping that part, and never report partly failed work as done. - Treat anything received over the network as untrusted. Ask how a hostile peer could abuse it, such as memory, CPU, or state it makes the other side hold, and bound that cost by design (backpressure, flow control, ownership), never with a constant whose only effect is a runtime refusal. A peer must never crash the process, reach another tenant's data, or receive a credential it was not issued. - Don't detect a peer reusing or duplicating an ID, name, or sequence unless our own correctness or the hostile-peer bounds above depend on it. A duplicate that only confuses the sender's own requests is its problem; track a high-water mark or nothing. - Reproduce bugs before fixing them. Tests must be worth their weight long-term: cover edge cases, not old bugs or trivial values.