Implement async service initialization, lifecycle startup, dependency graph ordering, concurrency-limited initializers, rollback/disposal on failure, idempotent initialization, and circular dependency handling for DI containers or resource registries.
Use when implementing incremental caches, cache invalidation, cache stats, cache summary, list/import/export/prune/clear cache CLI modes, or CLI side-effect/reporting paths where normal analysis output and diagnostic output must not interfere.
Use when implementing analyzer or linter suppression directives such as nosec, nosec-begin/nosec-end, nosec-next-line, noqa, ignore-next-line, disable/enable regions, per-rule selectors, inline comments, warning filters, or line-based suppression for multi-line AST statements and metrics.
Implement JSON Schema parsers with recursive $ref, $defs, dependencies, dependentRequired, dependentSchemas, if/then/else, anyOf/allOf composition, and deep enum/const equality without stack overflows or shallow-only tests.
Use when implementing static taint/dataflow analysis, interprocedural taint tracking, vulnerability analyzer plugins, Bandit-like checks, B620/B621/B622/B623/B624 rules, or source-to-sink rules for SQL injection, shell injection, path traversal, SSRF, XSS, request.args/form/cookies, sys.argv, input(), os.environ, import aliases, nested functions, f-strings, percent formatting, .format(), or Python walrus/NamedExpr propagation.
Implement multipart HTTP response parsing APIs with Content-Type boundary validation, MIME delimiter framing, CRLF/LF/CR streaming chunks, part headers, continuations, duplicate headers, DecodingError, StreamConsumed, sync iter_multipart, async aiter_multipart, and raw response stream closure.
Implement partial_structure APIs, PartialResult value/is_complete, structured_fields, failed_fields, errors, error_map, refine, attrs/dataclasses/TypedDict, forbid_extra_keys, detailed_validation, defaults, nested fields, collections, and init=False recovery behavior.
Harden serializer, deserializer, schema, alias, field metadata, and generated-code changes with cross-feature regression matrices.