Return the cached response, or None on miss / expired entry. An entry whose verification_timestamp is older than the TTL is a miss (t

Build a minimal passport JSON wrapping the seven aggregates the lint reads (six pre-v3.8.2 + uncited_audit_failures[]).

Assert lint exits non-zero AND stdout contains `invariant` tag.

Return a function that loads a YAML file into a Python dict.

Invoke a script via subprocess, capturing stdout/stderr as text. extra_env values are merged into os.environ (not replaced).

Return a list of violation strings; empty list = PASS.

Parse a draft (or, with ``fragment=True``, a patch ``new_text``). Fragment mode differences: no YAML-frontmatter detection (frontmatter is ne

Returns a fresh fully-valid reviewer_full contract. Callers may mutate freely.

Apply a single replace mutation to `repo_dir/rel_path`. Asserts the `old` string is present (so a refactored prompt that no longer matches sur

Validator with format_checker so format-keyword constraints (e.g. date-time on obtained_at) are actually enforced.

Return list of schema violation messages. Empty list means pass.

Smallest valid v3.6.4 entry; tests overlay trust fields on top.

Run `jq [-r] [-e] -f <filter>` over a Python payload. Returns (returncode, stdout).

Copy a fixture package into tmp and run the CLI on the copy. Returns (exit_code, report_dict, package_dir). The copy keeps the repo fixture p

Return {"matched": bool, ...}. Raise SemanticScholarUnavailable on transient API failures (after the protocol's retry budget is exhaus

Run the v3.6.8 lint as a subprocess (so its sys.exit propagates).

Load a JSON Schema file and verify it parses as Draft 2020-12. Centralised so test files don't re-implement the same load + check_schema pair

Run the lint script in `cwd`. Returns CompletedProcess.

Minimal SUPPORTED row — INV-1 positive baseline.

Return a function `_clean(doc, extra_blank=None)` that returns a copy of a dict/list with timestamp-like fields blanked to '<TIMESTAMP>'. By

Soft warnings per spec §4.3 (SC-1 baseline lag through SC-11 panel_size sanity). Non-blocking; printed to stderr by main().

Dotted-path get on a dict tree; returns None if any segment missing.

Return template text and assert baseline anchors are present.

Run the v3.7.1 audit-scope lint as a subprocess (so sys.exit propagates).

Returns a fresh fully-valid SR compliance report. Callers may mutate freely.

Everything lines up — baseline for PASS cases and drift mutations.

Parse the FIRST <!--ref:...--> marker found in `text`. Returns None if no ref marker is present (e.g. a bare `HIGH-BLOCK` token in prose — which m

Write report to a temp file and run the validator against it.

Minimal passport carrying the #260 aggregates the lint reads. A standard happy-path passport: one manifest with one experiment-backed claim,

Construct a Draft 2020-12 validator with format checking enabled. Without format_checker, format keywords (date-time on verified_at / acknowl

Copy the four lint-relevant files into a minimal repo tree under tmp_path.

Copy clean fixture to a tmp dir for mutation.

Reduce per-resolver outcomes to a 3-class lookup_verified value. Each value in `resolver_outcomes` is a dict with at least `status`, and (for

Title search under the #431 exact-title-or-bust gate. A candidate matches iff it clears the 0.70 ratio AND is an exact normalized tit

Build a clients dict of MagicMocks. Pass a configured mock per resolver, or None to get a default (all lookups miss).

Store (or overwrite) the resolver response, stamping it now (UTC).

Minimal UAF-INV-* positive baseline (v3.8.2 / #118).

Return the anchored document text (pure function; parser may raise).

Assert lint exits 0 (no findings) on a passport that should pass.

Build a synthetic claim_audit_result row for matrix-row tests.

Run the launcher with PATH = bin_dir + system dirs (bin_dir first, so injected interpreters shadow the real ones). Returns (exit_code, stdout, std

Validator with format_checker so format keywords (date-time on generated_at) are actually enforced. Same pattern as T2's test_literature_corpu

Return `(is_candidate, trigger_tokens)` for one sentence. Trigger tokens are returned in document order (left-to-right in the source sentence

Invoke the lint subprocess; returns (exit_code, stdout, stderr).

Verify one citation's existence across the four resolvers. `entry` carries citation_key, title, authors, year, source_pointer, optional doi /

Remove the exact (case-sensitive) literal; assert it was present so a drifted probe fails loudly instead of exercising nothing.

A well-formed experiment_provenance[] entry.

Run the CLI on a package dir; returns (exit_code, report_dict).

Run the v3.6.8 cite-provenance pipeline lint.