Repository navigation
ExecSurface v1.0.0
Status: PUBLISHED STABLE RELEASE — INTERNALLY QUALIFIED BOUNDED
ExecSurface v1.0.0 is the first stable release of the runtime behavioral verification / execution-semantics / semantic-evidence correctness line.
Published 2026-10-08 from source e70169b959f2163715090c371335fa6c5591e3e4.
Release evidence:
- GitHub Release workflow:
37771109825— SUCCESS; - crates.io publication workflow:
37771295572— SUCCESS; - stable Action channel:
AETHERXGLOBAL/execsurface@v1; - immutable Action pin:
AETHERXGLOBAL/execsurface@v1.0.0; - release asset SHA-256:
6892dc54e6f3f842cfecffa77ae0814c9e30aec219f2251c7c39a9f19f02ca7b; - exact registry install:
cargo install execsurface --version "=1.0.0" --locked— PASS.
Independent external validation is not claimed. The release is internally qualified under the repository's published compatibility, adversarial, artifact, rollback, productization, and release-control gates.
Supported boundary
The stable v1 support boundary is intentionally narrow:
- Linux x86_64;
- native
ptraceas the bounded public reference observer; - prebuilt release artifact qualified for the Ubuntu-24.04 / glibc-2.39-class environment;
- source / crates.io local-build path qualified on Ubuntu 22.04 and Ubuntu 24.04;
- declared Rust source floor: Rust 1.82.
This is not a claim of universal Linux compatibility, ARM64 support, eBPF authority, or backend interchangeability.
Stable verdict contract
ExecSurface v1 keeps the stable verdict/exit-code meanings:
- PASS =
0 - ERROR =
2 - REVIEW =
10 - BLOCK =
20
Incomplete, ambiguous, unsupported, or lost evidence cannot silently become PASS.
A baseline remains separate from policy. Backend identity alone does not create semantic authority.
Target-outcome boundary
The wrapped target command's native exit code or terminating signal remains report metadata rather than a drift-verdict input.
Therefore:
ExecSurface PASS does not mean the wrapped target command succeeded.
When target-command success matters, keep the target command as its own CI correctness gate.
Alpha.6 -> v1 compatibility
The internal RC qualification proved bidirectional profile-4 compatibility:
- an Alpha.6 profile-4 baseline can be consumed by the v1 candidate for unchanged PASS;
- controlled drift remains REVIEW/10;
- baseline bytes remain unchanged;
- a v1 profile-4 baseline can be consumed by Alpha.6 for unchanged PASS.
Alpha.5 profile-3 baselines remain immutable historical evidence and are explicitly incompatible with profile 4 rather than silently reinterpreted.
Policy compatibility
Policy schemas v1 and v2 retain their previous meanings.
Policy schema v3 remains opt-in and adds exact matching over already-qualified canonical evidence, including:
- path resolution;
- file-open intent;
- rename-source identity.
Unknown or unsupported schema states fail explicitly.
Semantic correctness work carried into v1
The stable release includes the Stage-2 remediation and post-R8 destruction results, including bounded corrections for:
- stale FD identity after dup-family replacement;
- silent untracked-FD I/O attribution;
- successful-open object/path authority;
- hardlink ambiguity;
- baseline and policy custody;
- policy expressiveness;
- CI trigger coverage;
- verdict-output aliasing and materialization;
- write/create/truncate/ftruncate/delete final-state ownership;
- rename lineage;
- symlink/hardlink output indirection;
- output-parent rebinding.
Negative and fail-first evidence is retained in repository history.
Release integrity
The v1 release transaction requires, in order:
- exact release request / package / immutable-tag identity;
- immutable v1 tag ruleset verification;
- full source gates;
- release binary build and smoke;
- checksum generation;
- GitHub provenance attestation;
- stable GitHub Release publication;
- zero-contact binary consumer proof;
- immutable tag source / Action proof;
- promotion of the moving
v1Action channel only after public proofs; - stable
@v1consumer proof; - crates.io publication only after the stable Action proof;
- exact-version registry install verification.
Performance boundary
Native ptrace has material runtime overhead on some workloads.
Internal fixed-workload RC characterization retained median ratios of approximately 4.3x–4.6x on the measured small workloads.
No universal low-overhead claim is made.
Research/product classification
ExecSurface v1 is a runtime behavioral verification / execution-semantics / semantic-evidence correctness product.
It is not classified as cybersecurity research.
It is also not antivirus, EDR, SIEM, malware detection, a sandbox, or proof that software is safe.