← ClaudeAtlas

no-silent-passlisted

Write Python whose failures are visible and checks that actually fire — distinct sentinels for success and error, filters that narrow to nothing without reading as "all clear", output never discarded on a non-zero exit, and selftest cases proven red before they are trusted (mutate the fix, mutate it the other way too, count occurrences rather than asserting presence, pick fixture values that tell the two versions apart). Use when adding or reviewing a `--selftest` case, when a guard or lint is added, when a function reduces observations to a number a threshold reads, when a bug is fixed and a regression case is written, or when a check has only ever been seen passing.
AleksandarBisevac/claude-plugins · ★ 4 · Code & Development · score 77
Install: claude install-skill AleksandarBisevac/claude-plugins
# No silent pass Two halves of one rule. Code must not produce normal-looking output when something went wrong, and a check must not look green while asserting nothing. Both fail the same way: everything reports fine and nobody learns anything for months. This repo's test suite is the `--selftest` block inside each file, printing `N/M cases passed`. Everything below is about those cases and the code they guard. ## Fail loud - **Give success and failure distinct sentinels.** If the error path writes the same value the success path writes, failed runs read as complete. `except Exception: status = "READY"` is the shape to look for. - **When you catch, either recover meaningfully or make the failure visible** — raise, or return a distinguishable value. A `return`/`continue`/fallback inside `except` that produces normal-shaped output is where silent corruption lives. - **`x = x or default` is a bug whenever `0`, `False` or `""` are meaningful.** Use `x if x is not None else default`. Reserve the `or` form for empty containers. - **A no-op on unexpected input is silent corruption.** Skipping a key you do not recognise, or `continue`-ing past a record you cannot parse, with no error and no count, leaves the caller believing the operation applied. - **Filtering down to empty must never read as "all clear."** When a selector, rule set or work list narrows to nothing, an "evaluated everything, found no problems" path reports a perfect result while checking nothing