Code that promises more than it can deliver, refused at check time.

A meaningful fraction of code-level "AI hallucination" has the same shape as a long-known software defect: a function declares it returns a complete answer about an input the program cannot fully process. furqan-lint makes that shape structurally rejectable on existing Python, Rust, and Go code -- no language migration required.

When a Python function returns Optional[X] and the caller silently narrows it to X, that is the structural equivalent of dropping the Incomplete arm of an Integrity | Incomplete union. The same shape appears as Result<T, E> in Rust and (T, error) in Go. Most checkers do not enforce this across language boundaries. furqan-lint does.

The core checks run on every file: D24 (all-paths-return), D11 (status-coverage), return_none_mismatch, and -- via furqan-lint diff -- additive_only across two versions of a module's public surface. Diagnostics name the rule violated, the producer, the consumer, and the minimal fix.

The same checker substrate runs on Python, Rust, and Go. Each check names a structural failure mode that has appeared in published incident reports and in the audit chain that backs this release.

The checker emits one of four verdicts per module. The first three carry diagnostics. Only PASS is silent.

v1.0.0 closes a canonical chain that started at G10.5 al-Mubin (v0.11.1) and ran through at-Tawbah, al-Hujurat, as-Saff, al-Mursalat, an-Naziat (v0.13.0), and Tasdiq al-Bayan (v0.14.0). The closing move: furqan-lint signs its own release with Sigstore via ambient OIDC, publishes the signed self-manifest as a Release asset, and exposes Relying-Party verification through a CLI subcommand.

The self-manifest pins the source list of 24 substrate modules (4 top-level checker substrate plus 20 gate11 substrate including two new self-attestation modules) and binds them with a Form A checker_set_hash. There are two of these: a universal hash over all checker substrate, and a self-attestation hash over the pinned-self subset. Both are published in SAFETY_INVARIANTS.md and verified at install time.

On the relying-party side, after pip install:

A drift in any of the pinned source files, or a missing self-manifest, or an unexpected signer identity, surfaces as CASM-V-072 (self-attestation-failure) with one of three named sub-conditions: manifest-not-found, checker-set-hash-drift, or signature-verification-unexpected. The 732-test substrate (687 passing, 45 environment-gated) covers the verification path end-to-end.

Honest asymmetries. Self-attestation ships on the Python distribution. The Rust verifier substrate is part of the Python wheel rather than a separate crate; the Go self-attestation path is deferred. The published docs/gate11-symmetry.md row records both.

The four checks that ship in furqan-lint (D24, D11, return_none_mismatch, additive_only) -- plus R3 on the Rust adapter -- are not bespoke heuristics. They are translations of compile-time primitives from Furqan, the research programming language whose multi-checker pipeline (bismillah scope, zahir/batin, mizan calibration, tanzil build ordering, ring-close, scan-incomplete, status-coverage, return-type-match, all-paths-return) defines the structural-honesty discipline.

The language gives the discipline; furqan-lint carries it into existing Python, Rust, and Go code -- no migration required. You do not need to install Furqan to use furqan-lint. The architecture-derivation paper is listed under Research below; the Furqan repository is linked for researchers who want the substrate directly.

furqan-lint is the production surface of a published research program. The Munafiq Protocol names the failure mode RLHF and Constitutional AI do not address: a system can be Compliant without being Aligned. The Furqan language turns that protocol into compile-time checks, and furqan-lint carries those checks into Python, Rust, and Go. The cross-text empirical paper grounds both in measured structural properties of the source corpus, not in metaphor. Every paper carries a permanent Zenodo DOI and an auditable claim against a named null hypothesis.