This document defines how WrightKit tests components against hostile or pathological input. Load it when a change affects how a parser, compiler, analyzer, language service, protocol handler, or agent-facing tool handles malformed, extreme, or adversarial input, or when adding fuzzing.
It is part of the testing policy, whose core principles also apply, including observable failures and negative-path coverage.
Parsers, compilers, analyzers, language services, protocol handlers, and agent-facing tools should be tested against hostile or pathological input where relevant. Useful classes include:
- deep nesting;
- unexpected end-of-file at syntax boundaries;
- recursive or cyclic imports/includes;
- recursive macro expansion;
- large arrays or strings;
- extreme numeric literals;
- Unicode and encoding boundaries;
- duplicate declarations or identities;
- missing mappings or catalog entries;
- malformed external/process responses;
- resource limits and interruption behavior.
User-controlled input must not cause an uncontrolled panic, abort, hang, or fabricated success result. Explicitly documented resource exhaustion or unsupported behavior is acceptable when surfaced deterministically.
Fuzzing is encouraged for parsers, serializers, protocol boundaries, and other high-input-space components when it adds meaningful coverage beyond hand-written cases.