Repository navigation
Releases
Invaris v0.5.1
Invaris v0.5.1
- Run a full suite and select its reported test IDs with repeatable
test-result --test-id ID. Missing selected tests produce an inconclusive result. - Use
validity-result tscto distinguish recognized TypeScript source errors from compiler setup failures during mutation validation. - Use per-action help and read failed mutation stage details in
explain. - Follow the refreshed documentation homepage for installation and a runnable guard challenge.
Existing harnesses and default evidence versions remain supported. The explanation schema adds an optional details field. Review installation and the harness reference.
Invaris v0.5.0
Invaris v0.5.0
Invaris 0.5.0 closes the gaps found when Invaris was used on a real Node project: it is easier to discover, mutation harnesses can delete a line, common test reporters plug into mutation runs without a hand-written adapter, and guards can report richer findings.
invaris --help(also-handhelp) prints usage and exits0, andinvaris --version(alsoversion) prints the version.- Mutation (v1alpha3) harnesses accept
"operator": "delete", which removes the single occurrence ofbefore; omitafteror set it to"". See mutation operators andexamples/mutation/deleted.json. invaris test-result <go-test-json|jest-json|junit-xml> [FILE|-]convertsgo test -json, jest or vitest JSON, and pytest JUnit XML into theinvaris.test-result/v1alpha1object a mutation test command prints. It exits0when every test passed,1when an assertion failed and2when the run failed outside an assertion: a build failure, a timeout,TestMainexiting nonzero, a suite that failed to load or whoseafterAllthrew, or a pytest setup error. Exit2is an infrastructure failure, so it never counts as a killed mutant. Skipped tests are left out, so a skipped selected test makes the experiment inconclusive. See test-result converters.- Structured inspection findings accept optional
ruleId,messageandline. Reports carry the rule and an optionalline; the v1alpha5 portable profile still withholds messages. A challenge proof for a structured command guard matches the finding's policy and path, whateverruleIdthe guard reports. See inspection. - Campaign reports record the paths that harness
workspaceblocks left out of the snapshot (sourceIdentity.workspaceExclude) and a raised entry bound (workspaceMaxEntries), andinvaris explainreports them as theworkspace-excludedlimit. - Ecosystem publishes the Evaris compatibility table.
Default inputs, report versions and verification exit codes are unchanged, and the new fields are optional and omitted when unused. Consumers that strictly decode findings must accept the optional line before reading reports from guards that emit it: Evaris 0.4.0 accepts it, while Evaris 0.3.0 and earlier refuse such reports, so upgrade Evaris first or together with Invaris. A harness that uses delete needs 0.5.0; earlier binaries refuse it. Reports are not authenticated: matching hashes show that bytes are consistent, not who ran a check or how.
Known limitations:
test-resultsupports three reporter formats. Test IDs follow each reporter's naming (package.Test,file::full name,classname::name), sotestIdsmust use the converted IDs. Write jest or JUnit reports to stdout or to a path excluded byworkspace.exclude, so the report file is not a workspace change.- Excluded paths and dependencies remain outside the integrity check; campaign reports now list them, but a command can still change them without the run noticing.
- The self-check experiment has not been repeated with these features or by a second operator.
Linux and macOS archives for amd64 and arm64 include both invaris and invaris-command-provider, a README and the Apache-2.0 license. Verify downloads with SHA256SUMS. No Go installation is needed to run the binaries; source builds require Go 1.27.4 or later. Windows is supported through WSL using the Linux archives.
Commands run with your user permissions and can access the network and external files. Disposable copies are not security sandboxes.
See the changelog, installation guide, evidence reference and threat model.
Invaris v0.4.0
Invaris v0.4.0
Invaris 0.4.0 helps a developer run a first useful experiment and read the result: invaris doctor finds setup problems before a run, invaris explain says what a saved report does and does not show, and a harness can now run in a Node project with node_modules installed.
invaris initcreates.invaris/invaris.json; commands prefer that file and fall back to the root-level configuration when it is absent. Existing root-level files remain supported, andinitrefuses to create a competing configuration. Campaign configuration can also live in.invaris; see authoring and campaigns.- An
invaris-clinpm package is prepared for publication, containing the CLI and command provider for Linux and macOS on amd64 and arm64. npm installation requires Node.js 22.14 or later; see the installation guide for current availability. invaris doctor [HARNESS]inspects a harness without running any of its commands. It checks that the harness loads and which actions it supports, that the directory can be copied into a workspace (the check that names an installednode_modulesor a symbolic link), that each program is found in the PATH its command would use, allowlisted variables that are unset, that targets exist, that literal edit markers occur exactly once, that built-in operators apply and pinned preimages still match, and that writable paths exist. Provider discovery is reported as skipped because it would run the adapter. Exit0means nothing failed and1means a setup check failed.invaris explain REPORTreads a v1alpha2 to v1alpha5 report offline, validates it, and labels each resultdetected,missed,inconclusive,excluded,passed,failedornot-run, with the meaning and next step of its recorded reason. It lists limits such as weak marker evidence, a portable profile that withholds details, a dirty working tree and an incomplete decision. A report that fails validation, such as one whose decision was edited, is refused.- Mutation (v1alpha3) and challenge (v1alpha4) harnesses accept an optional
workspaceblock.dependencieslinks top-level directories such asnode_modulesbeside the disposable copy instead of copying them, so Node and Vite resolve packages while the copy holds no link into the source.excludeleaves build output and tool caches such as.nextout of the copy and its integrity snapshots.maxEntriesraises the 2000-entry bound up to 200000. A campaign combines the blocks of the harnesses it references. See workspace options. - A
workspace_changedresult now names what changed, up to 20 paths marked+(added),-(removed) or~(changed), such as+node_modules/.vite/deps. The entry-limit and symbolic-link refusals say whichworkspacefield resolves them. - Both commands print text by default and JSON with
--format json; the JSON is described by new schemas. Neither command writes a report or runs project code. - A self-check experiment shows the whole workflow on this repository: a healthy baseline, a detected fault, a missed-detection control, a killed and a surviving mutant, and a campaign, with a friction log and what
doctorandexplainchanged.
Existing root-level harnesses, report versions and verification exit codes remain supported; the default location created by init and configuration discovery have changed as described above. A harness that uses the workspace block needs 0.4.0; earlier binaries refuse it as an unknown field rather than ignore it. Reports are not authenticated: matching hashes show that bytes are consistent, not who ran a check or how.
Known limitations:
doctorcannot tell whether a command prints the result object its role requires, or whethergo run ./toolcompiles, because that would mean running it. It inspects one harness at a time; run it for each file a campaign names.- Without a
workspaceblock, a checkout withnode_modulesinstalled still cannot be copied: the default copier refuses symbolic links and more than 2000 entries, anddoctornames the offending path.dependenciesaccepts only top-level directories; a nestednode_modulescan be excluded but not linked. - Excluded paths and dependencies are outside the integrity check. A command can change them, or write through a dependency link into the source, without the run noticing. The harness digest in each report pins the options, but a campaign report's
sourceIdentity.excludedstill lists only.gitand.invaris. - The self-check experiment below was run before the
workspaceblock existed and has not been repeated with it. - Two smaller friction items from the self-check are unchanged:
--reportrefuses to overwrite an existing file, and a challenge with a missed control exits1;explainshows the decision but does not explain the exit code. - The self-check has not yet been repeated by a second operator, which the roadmap asks for before claiming that the workflow generalises.
Linux and macOS archives for amd64 and arm64 include both invaris and invaris-command-provider, a README and the Apache-2.0 license. Verify downloads with SHA256SUMS. No Go installation is needed to run the binaries; source builds require Go 1.27.4 or later. Windows is supported through WSL using the Linux archives.
Commands run with your user permissions and can access the network and external files. Disposable copies are not security sandboxes.
See the changelog, diagnostics, installation guide, evidence reference and threat model.
Invaris v0.3.0
Invaris v0.3.0
Invaris 0.3.0 makes harnesses easier to create and upgrade, adds three deterministic guard-challenge operators, and provides portable evidence for sharing results.
invaris initdetects project markers and creates a generic or Go starter harness without running project code or overwriting a file.invaris migrateupgrades v1alpha1 harnesses to v1alpha2 while preserving commands and effective gates. Preview with--dry-run; a write keeps an exact.bakfile.checkandchallengesupport--format text|compact; campaigns also support compact output. JSON remains the default and report files remain JSON.- v1alpha4 guard harnesses can name
trailing-whitespace,crlf-injectionandsecret-leak-injection, pinned to an exact preimage. Built-in edits require structured detection evidence. - v1alpha5 evidence adds mutation metadata, reproduction identities, command digests and campaign effectiveness rows while withholding machine paths, command arguments, output and free-text details. It is automatic for v1alpha4 harnesses and opt-in for older inputs with
--evidence-version v1alpha5. - Workspace preparation now stops on cancellation, removes partial copies and records cancellation or timeout as the outcome reason.
Existing harnesses retain their default report versions and exit codes. A v1alpha4 harness needs a 0.3.0 binary; older binaries refuse it. Migration changes the config digest and reports compatibility as native. The fixed synthetic secret is a test fixture, not a working credential, and requires a scanner rule that recognizes it.
Linux and macOS archives for amd64 and arm64 include both invaris and invaris-command-provider, a README and the Apache-2.0 license. Verify downloads with SHA256SUMS. No Go installation is needed to run the binaries; source builds require Go 1.27.4 or later. Windows is supported through WSL using the Linux archives.
Commands run with your user permissions and can access the network and external files. Disposable copies are not security sandboxes. IDs and configured target names remain visible in portable evidence; neither hashes nor reports authenticate who ran a check.
See the changelog, installation guide, authoring guide, operator catalog, evidence reference and threat model.
Invaris v0.2.0
Invaris v0.2.0 adds campaigns. A campaign runs several mutation and guardrail experiments together from one frozen copy of the source, repeats them, and decides on the whole set with an explicit acceptance policy.
- Run
invaris campaignwith aninvaris.campaign/v1alpha1file that references existing v1alpha3 mutation and v1alpha2 challenge harnesses. Reports useinvaris.evidence/v1alpha4. - Repeats keep every attempt and are classified CONSISTENT, MIXED, or INCOMPLETE, with no voting.
- The mutation score and the guardrail proof rate count consistent experiments once each, are
nullwhen nothing is eligible, and are never combined. - A typed acceptance policy decides ACCEPT (exit 0), REJECT (1), or REFUSE (2). Incomplete or contradictory evidence refuses before any reject.
--onlyruns a recorded subset,--format textprints a summary, and--parallel Nruns attempts concurrently (default 1). Harnesses may sit in subdirectories, and optionaltoolsrecord tool versions.- Archives include the CLI and generic command provider. Verify downloads with SHA256SUMS.
Compatibility: run, check, challenge, and mutate keep their inputs, exit codes, and report versions. The v1alpha2 and v1alpha3 evidence schemas now require the reason and provenance keys added since 0.1.0. In challenge, a guard that writes into its workspace now yields INCONCLUSIVE, and rejected structured output from a guard exiting 0 or 1 is reported as malformed_evidence. See the changelog and the Campaigns guide.
Commands and evaluators run with the user's permissions. Disposable workspaces and campaign snapshots are not security sandboxes. See the threat model.
Invaris v0.1.0
Invaris v0.1.0 provides trusted local workflow execution, guard challenges, and single-mutant testing with structured JSON evidence.
- Run
invaris mutatewith a v1alpha3 harness to evaluate a single replacement against a baseline, validity check, and selected tests. - Mutation verdicts distinguish KILLED, SURVIVED, INVALID, and INCONCLUSIVE.
- Guard challenges match the expected policy, rule, and target and record cleanup and mutation digests.
- Archives include the CLI and generic command provider. Verify downloads with SHA256SUMS.
Existing v1alpha1 and v1alpha2 harnesses emit v1alpha2 evidence. See the evidence migration guide.
Commands and evaluators run with the user's permissions. Disposable workspaces are not security sandboxes. See the threat model and changelog.