Topics
One URL per cluster. The H1 is the question they typed.
Discovery pages, dated. Newest first in the log; here they sit with the day they shipped. If an existing section already owns the cluster, we edit that section instead of minting a twin.
-
Solver latency clean
A last VALID on
realize specs/system/authafter four minutes is not a cheap proof.proof audit --check solver_latency_cleannames the silent slice. Z3/Kind2 on one function is not that hop. Jama still authors. -
Trivial lemma
A last PROVE on
AddModel(a, b) == a + bcan be a restatement of the model body.proof audit --check trivial_lemmanames the silent lemma. Lemma branch coverage is not that hop. Jama still authors. -
Lemma branch coverage
A last PROVE on the positive arm of Abs can leave the negative arm with an empty
covered_by.proof audit --check lemma_branch_coveragenames the silent lemma. Lemma binding freshness is not that hop. Jama still authors. -
Lemma binding freshness
A separate-file lemma can keep its last PROVE after the bound function moved.
proof audit --check lemma_binding_freshnessnames the silent binding. Formalization lemma verdict consistency is not that hop. Jama still authors. -
Formalization lemma verdict consistency
A SYS-REQ can keep
formalization_status: validafter the lemma cache is still timeout.proof audit --check formalization_lemma_verdict_consistencynames the silent overstatement. Z3/Kind2 on one function is not that hop. Jama still authors. -
Approved guarantee KI conflict
An approved guarantee can keep its stamp after an open known issue names the same id.
proof audit --check approved_guarantee_ki_conflictnames the silent approval. Approvals current is not that hop. Jama still authors. -
Contract alignment clean
A cobra file can keep growing flags after the help topic never landed.
proof audit --check contract_alignment_cleannames the silent command. Design by contract is not that hop. Jama still authors. -
Assume contract consistency
A host can keep
// reqproof:assume ClearSessionafter the callee lost its lemma.proof audit --check assume_contract_consistencynames the silent assume. Design by contract is not that hop. Jama still authors. -
Annotation validity
A comment can still name
SYS-REQ-9999after the YAML is gone.proof audit --check annotation_validitynames the silent ID. Autolink clean is not that hop. Jama still authors. -
Gaps clean
An output can stay
?while Kind2 still prints realisable.proof audit --check gaps_cleannames the silent unconstrained output. Variable drift is not that hop. Jama still authors. -
Circular deps clean
Two files can parent each other while coverage still prints 100%.
proof audit --check circular_deps_cleannames the silent loop. Webpack imports are not that hop. Jama still authors. -
Role references resolve
A role can name a renamed check, a deleted help topic, or a retired requirement while the playbook stays green.
proof audit --check role_references_resolvenames the silent drop. Empty-is-win language is not that hop. Jama still authors. -
Autolink clean
A comment can name an unconfigured prefix and still vanish from the graph.
proof audit --check autolink_cleannames the silent drop. Zero annotations still pass. Jama still authors. -
Role prompt hygiene
A role can prize an empty finding list, or cap findings per worker, while the sweep stays green.
proof audit --check role_prompt_hygienenames the silent empty. A resource cap is not that hop. Jama still authors. -
Orphan code
A production helper can ship with no
Implements:line while coverage stays green.proof lint --check orphan_code_cleannames the silent shrink. Dead code is not that hop. Jama still authors. -
Code signal deadline cast unreviewed
An
executeAfterrow packed intouint32can sit as cast noise with no packing campaign.proof audit --check code_signal_deadline_cast_unreviewednames the silent packing. CodeSignal is not that hop. Jama still authors. -
Code signal suppressions reviewed
An
http.*skip can sit with no reason, an expired date, and no path.proof audit --check code_signal_suppressions_reviewednames the silent exception. CodeSignal is not that hop. Jama still authors. -
Code signal obligations reviewed
A SARIF hit can sit on FetchPolicy while the owner never listed the timeout class.
proof audit --check code_signal_obligations_reviewednames the silent scanner. CodeSignal is not that hop. Jama still authors. -
Obligation witness grounded
A coverage-trackable colon triple can sit on a tautological test that never runs the impl.
proof audit --check obligation_witness_groundednames the silent theater. Jama still authors. -
Obligation profile evidence complete
A class mapped to an evidence profile can sit with a colon triple and no passing result file.
proof audit --check obligation_profile_evidence_completenames the silent triple. Jama still authors. -
Obligation suppression reviewer
An error-severity suppress can sit with a long reason and no
reviewed_by.proof audit --check obligation_suppression_reviewernames the silent unsigned skip. Jama still authors. -
Obligation suppression rationale
A suppress can sit with reason TODO.
proof audit --check obligation_suppression_rationalenames the silent placeholder. Jama still authors. -
Obligation delegation resolves
A delegated_to id can sit after the target dropped the class.
proof audit --check obligation_delegation_resolvesnames the silent pointer. Jama still authors. -
Obligation decomposition complete
A parent can list a class and the satisfying child never carries it.
proof audit --check obligation_decomposition_completenames the silent drop. Jama still authors. -
Obligation baseline
Tags can fire a catalog class with no accept and no suppress.
proof audit --check obligation_baselinenames the silent omission. Jama still authors. -
Obligation enforcement backed
A cataloged checklist class can sit with no signal and no evidence.
proof audit --check obligation_enforcement_backednames the silent no-op. Jama still authors. -
Acceptance witness quality
An
:acceptancetag can sit on a child unit test.proof audit --check acceptance_witness_qualitynames it. Jama still authors. -
Acceptance review current
An AC list can sit on the STK with no stamp.
proof audit --check spec_lint_acceptance_review_currentnames it. Jama still authors. -
Concern adjudicated
An open YAML note can sit forever with no review_due.
proof audit --check concern_adjudicatednames it. Jama still authors. -
Problem reports reviewed
A DEFECT can stamp covered while the hardening never landed.
proof audit --check problem_reports_reviewednames it. Jama still authors. -
Change record lands
A CHG can close with an affects list while no requirement cites it back.
proof audit --check change_record_landsnames it. Jama still authors. -
Approval motivation present
A second stamp that still matches the shall looks current if the fingerprint did not change.
proof audit --check approval_motivation_presentnames it. Jama still authors. -
Mirror stale
A source edit that only duplicates a Verifies line still looks complete if the annotation cells did not move.
proof audit --check mirror_stalenames it. Jama still authors. -
Catalog semantic bindings reviewed
A scan match that is neither bound nor rejected still looks reviewed if catalog completeness stayed green.
proof audit --check catalog_semantic_bindings_reviewednames it. Jama still authors. -
Defect review current
A spec with code links and no
defect_reviewstamp still looks reviewed if no KnownIssue was filed yet.proof audit --check defect_review_currentnames it. Jama still authors. -
Severity inflation
A
criticalstamp with no surface, no vector, and no risk basis still looks like disclosure-grade work.proof known-issue checknames it. Jama still authors. -
CVSS severity consistent
A
lowlabel on a still-critical CVSS vector drops the record out of every high-end hop.proof audit --check cvss_severity_consistentnames it. Jama still authors. -
Security relevance consistent
A security-named KnownIssue with
cve_surface: noneand no CVSS vector drops out of the security lens.proof audit --check security_relevance_consistentnames it. Jama still authors. -
MC/DC ignore classified
A green coverage cell on a bare ignore is bookkeeping, not a named exemption.
proof audit --check mcdc_ignore_classifiednames it. Jama still authors. -
MC/DC known issue disposition stale
A KI-gated exemption that outlives its fixed bug is a leftover lie.
proof audit --check mcdc_known_issue_disposition_stalenames it. Jama still authors. -
Known issue template transfer
An open KnownIssue that names a class without listing isomorphic sites leaves the second sink untracked.
proof audit --check known_issue_template_transfernames it. Jama still authors. -
Known issue sibling disposition
A fixed KnownIssue that leaves open siblings or unmarked sites is a silent close.
proof audit --check known_issue_sibling_dispositionnames it. Jama still authors. -
Known issue severity prose consistent
A title that reads HIGH on a record filed low is a quiet downgrade.
proof audit --check known_issue_severity_prose_consistentnames it. Jama still authors. -
Known issue security remediation present
A security-relevant KnownIssue with no remediation or mitigation is unnamed containment.
proof audit --check known_issue_security_remediation_presentnames it. Jama still authors. -
Known issue recheck due
A disclosed KnownIssue whose recheck date has lapsed is overdue.
proof audit --check known_issue_recheck_duenames it. Jama still authors. -
Deployment tvl verified
A high claim with no TVL and no outside grade is unanchored.
proof audit --check deployment_tvl_verifiednames it. Jama still authors. -
Submission validated
A submitting KnownIssue with empty gates is not ready.
proof audit --check submission_validatednames it. Jama still authors. -
Poc quality checked
A missing poc_quality block is not a review.
proof audit --check poc_quality_checkednames it. Jama still authors. -
Known issue poc quality effective
A present but false poc_quality block is not a review.
proof audit --check known_issue_poc_quality_effectivenames it. Jama still authors. -
Known issue mirror anchor relevant
A declared mirror: whose resolved scope the KI never touches is off-scope.
proof audit --check known_issue_mirror_anchor_relevantnames it. Jama still authors. -
Known issue affected requirements present
An empty affected_requirements list is invisible to the trace graph.
proof audit --check known_issue_affected_requirements_presentnames a contract-less KI. Jama still authors. -
Known issue reproducer consistent
An accept-verb test name is not the secure contract.
proof audit --check known_issue_reproducer_consistentnames a contradiction. Jama still authors. -
Known issue reproducer present and resolves
Prose in reproduction_steps is not a runnable selector.
proof audit --check known_issue_reproducer_present_and_resolvesnames a stale command. Jama still authors. -
High severity reproducer grade
A CVSS high is not an executed witness.
proof audit --check high_severity_reproducer_gradenames a silent headline. Jama still authors. -
High severity obligation without vector
A CVSS high is not a campaignized class.
proof audit --check high_severity_obligation_without_vectornames a silent hazard. Jama still authors. -
Vector campaign closure
An ad campaign is not an opened hunt with an exit.
proof audit --check vector_campaign_closurenames a silent partial. Jama still authors. -
Obligation evidence complete
A green checklist is not a triple.
proof audit --check obligation_evidence_completenames a missing kind. Jama still authors. -
Code signal unbindable
A CodeSignal interview is not a draft class.
proof audit --check code_signal_unbindablenames a silent proposal. Jama still authors. -
Vector campaign hygiene
An ad campaign is not a hunt YAML.
proof audit --check vector_campaign_hygienenames a dangling KI. Jama still authors. -
Unresearched P0 vectors
A CodeSignal interview is not an opened hunt.
proof audit --check unresearched_p0_vectorsnames a silent P0. Jama still authors. -
Known issue complete
A Salesforce bulletin is not a failing test.
proof audit --check known_issue_completenames a sticky note. Jama still authors. -
Data constraints complete
A SQL CHECK is not a partition.
proof audit --check data_constraints_completenames a gap or an overlap. Jama still authors. -
Catalog version pinned
A Magento SKU is not a catalog pin.
proof audit --check catalog_version_pinnednames a pin that no longer matches. Jama still authors. -
Change evidence complete
A Jira ticket is not a covering DEFECT.
proof audit --check change_evidence_completenames a declared fix without backing. Jama still authors. -
Residual kill hygiene
A spreadsheet kill is not a sample_table.
proof audit --check residual_kill_hygienenames a closed kill without samples. Jama still authors. -
Approvals current
A Slack stamp is not a current fingerprint.
proof audit --check approvals_currentnames a stale shall. Jama still authors. -
Known issues reviewed
A Salesforce bulletin is not a dated YAML.
proof audit --check known_issues_reviewednames a stale review_date. Jama still authors. -
Waivers reviewed
An empty expiry is not a review.
proof audit --check waivers_reviewednames the expired skip. Jama still authors. -
Documented claim verified
A Sphinx percent is not a named claim cell.
proof audit --check documented_claim_verifiedwarns when the cell is incomplete. Jama still authors. -
Variable drift
A green gaps report is not list/sentence agreement.
proof audit --check variable_driftnames the requirement. Jama still authors. -
Vacuous requirements
A green SAT is not a constraint.
proof audit --check vacuity_cleannames the vacuous or still-unchecked requirement. Jama still authors. -
Obligation completeness
A green STK is not a covered checklist.
proof audit --check obligation_completenessnames the class with no SYS-REQ. Jama still authors. -
Verification chain
A green parent is not a verified tree.
proof gaps specs/system --check verification_chainnames the child that never ran. Jama still authors. -
Orphan tests
A green Test* is not evidence until it names a live requirement.
proof lint --check orphan_tests_cleannames the symbol. Jama still authors. -
Catalog completeness
Four invariants on yaml, catalog, rules, and checklists.
proof audit --check catalog_completenessfails the drift. Jama still authors. -
Software verification
A green suite is not the graph pipeline.
proof verifyruns it. Jama still authors. -
Ambiguous requirements
Two green shalls can still share a variable.
proof audit --check ambiguity_reviewednames the pair. Jama still authors. -
Spec conformance
A green formula is not a read of the Go.
proof audit --check spec_lint_spec_conformance_review_groundedwants a citation. Jama still authors. -
Non functional requirements
A green quality list is not a category table.
proof statusnames the mix. Jama still authors. -
Software metrics
A green DORA board is not spec-set health.
proof gaps specs/system --check metricsnames the pile. Jama still authors. -
Software checklist
A green quality PDF is not a campaign stamp.
proof audit --check process_checklistnames the next eligible step. Jama still authors. -
Taint analysis
A green sanitizer list is not a trust fact.
proof audit --check trusted_outputs_from_untrusted_inputsreadstrust: client. Direct edges only. Jama still authors. -
Test impact analysis
A green full suite is not a named plan.
proof test affectedreads traces and Verifies rows. Incomplete evidence falls back to the full suite. Jama still authors. -
Requirement review
A Slack walkthrough is not a brief.
proof review req SYS-REQ-010names traces, witnesses, and impact. It does not move status. Jama still authors. -
Requirements validation
A silent drop is not valid.
proof validatefails the unquoted colon. SEBoK still owns Boehm. Jama still authors. -
Requirements quality
A green description is not quality.
proof gaps specs/system --check qualityfails empty FRETish. Jama still scores the programme. Jama still authors. -
Requirement lifecycle
A Slack done is not review.
proof req status SYS-REQ-010 --to reviewrecords the hop. BABOK still owns the knowledge area. Jama still authors. -
Stakeholder requirements
A SYS-REQ import is not L0.
proof audit --check stakeholder_requirements_existfails whenspecs/stakeholderis empty. SEBoK still owns the workshop. Jama still authors. -
Coverage threshold
A global 82% is not the changed shall.
proof audit --check coverage_thresholdjudges each in-scope requirement against an imported profile. Vitest still owns the runner. Jama still authors. -
Equivalence partitioning
A wiki of valid vs invalid is not a domain.
proof verify-properties specs/systemasks Z3 whether authored partitions cover the input without overlap. GeeksforGeeks still owns the glossary. Jama still authors. -
Slow tests
A suite wall clock is not a named leaf.
proof test slowlists cases whose body crossed the threshold. pytest-benchmark still owns the timer. Jama still authors. -
Documentation coverage
A Sphinx percent is not a requirement doc.
proof audit --check documentation_coverageinventoriesdocumented_bypaths. Compodoc still owns the docstring bar. Jama still authors. -
Requirements assumptions
A Slack “we assume the vendor is up” is not a boundary.
proof req assumptions listinventoriesreq_type=assumptionwith an owner and a review date. PMI still owns the project log. Jama still authors. -
Formal methods
Three green fixtures are not a universal claim.
proof verify-properties specs/systemasks Z3 on authored variable properties. Coq still owns the assistant. Jama still authors. Proof does not infer the property. -
What is model checking, and how do I keep the traces true in CI?
A lecture is not a verdict.
proof realize specs/system autopilot --diagnoseasks Kind2 on the Lustre contract. SPIN still owns Promela. Jama still authors. Proof does not implement the checker. -
Risk acceptance
A Slack “we accept this” is not a signature.
proof risk acceptrefuses a silent write without--accepted-byand--review-date. Metricstream still owns GRC. Jama still authors. -
Loop invariant
A green walk of the slice is not establishment.
proof verify-lemma --solver z3,cvc5 --tags reqproof_proof ./...refuses a silent pass when aforhas no// reqproof:invariant, or when preservation is SAT. Dafny still owns the language. Jama still authors. -
Negative testing
A green authorized request is not a reject.
proof audit --check negative_path_witness_requiredrefuses a silent pass when a security-classed shall has only a happy-path test. go test still runs the case. Jama still authors. -
Integration testing
Green children are not the wired boundary.
proof audit --check integration_evidence_witnessedrefuses a silent pass when an INT-REQ has only unit tests. Testcontainers still runs the stack. Jama still authors. -
Fuzz testing
Invalid inputs are the fuzzer's job. A
:fuzztriple on aTest*function is not.proof audit --check fuzz_evidence_carrier_validrefuses annotation-only credit. go test -fuzz still mutates. Jama still authors. -
Acceptance criteria
A Done list on a user story is not an acceptance test.
proof audit --check acceptance_criteria_witnessedrefuses a silent pass when a child's unit tests stand in for the assembled whole. Jira still owns the story. Jama still authors. -
Software problem report
Class-closure after the fix, not a Done ticket.
proof problem-report validaterefuses a closed stamp without hardening. Jira still owns the tracker. Jama still authors. -
Pull request review
The requirement blast radius on this branch, not LGTM on the Go.
proof review-prlists changed shalls, approval drift, and files to inspect. GitHub still owns the thread. Jama still authors. -
System requirements review
The NASA first gate from this graph this commit, not last month's SRR pack.
proof gate srrassesses existence, validation, levels, draft status, and blocking risks. Jama still authors. -
Critical design review
The NASA detailed-design gate from this graph this commit, not last month's CDR pack.
proof gate cdrassesses annotations, autolink, the build, and orphan code. Jama still authors. -
Test readiness review
The NASA test gate from this graph this commit, not last month's TRR pack.
proof gate trrassesses tests, coverage, MC/DC, fixtures, Z3, and suspect links. Jama still authors. -
Software verification report
The completeness dump from this graph this commit, not last quarter's V&V PDF.
proof doc generate verification-reportprints the current checks. Jama still authors. -
Interface control document
The INT-REQ that still matches both sides this commit, not a PDF from last quarter.
proof audit --check interface_staleness_cleanfails when the fingerprint drifted. Jama still authors. -
Requirements diagram
The live SpecTree as Mermaid, not a SysML drawing from last quarter.
proof diagram hierarchy --format mermaidprints the counts. Jama still authors. -
Requirements decomposition
One parent shall split into owned software or interface children.
proof req decomposepreviews the files;spec_lint_decomposition_adds_refinementfails a copy-paste child. Jama still authors. -
Inconsistent requirements
Two shalls that cannot both hold.
proof check consistencyfails on UNSAT;consistency_pair_coveragefails when zero pairs were examined. Jama still authors. -
What is a software verification plan, and how do I keep it true in CI?
Per-requirement strategies and expected evidence, before the code.
proof verify-plan status --checkfails when a planned item does not resolve. Jama still authors. -
What is requirements completeness, and how do I keep it true in CI?
Declared classes, catalog entries, signal rules, and checklists.
proof audit --check catalog_completenessfails when they disagree. Jama still authors. -
What is software change impact analysis, and how do I keep the blast radius true in CI?
The blast radius of a change.
proof trace impactprints the graph. Jama still authors. LDRA still wins at C CIA. -
What is ARP 4754A, and how do I keep the system requirements true in CI?
SAE guidance for civil aircraft and systems.
proof req importlands the allocation. Jama still authors. Proof does not run FHA or PSSA. -
What is SARIF, and how do I keep the findings true in CI?
OASIS JSON for analyzer findings.
proof signals importbinds the row. Jama still authors. Proof is not a SARIF viewer. -
What is k-induction, and how do I keep the traces true in CI?
Kind2's base-plus-step engine on a Lustre contract.
proof realizeasks Kind2. Jama still authors. Proof does not implement k-induction. -
What is design by contract, and how do I keep the contracts true in CI?
Assume/guarantee at a component boundary.
proof check integrationnames an unmatched assume. Jama still authors. Proof is not Eiffel. -
What is ASD-STE100, and how do I keep the prose true in CI?
Simplified Technical English as a lint on the YAML.
proof lint --check spec_lint_prose_ste100names the hedge. Jama still authors. Proof is not an ASD trainer. -
What are derived requirements, and how do I keep them true in CI?
A shall that was not in the stakeholder set, filed as a SYS-REQ.
proof req derivewrites the YAML and the satisfies link. Jama still authors. Proof does not invent architecture constraints. -
What is linear temporal logic, and how do I keep the traces true in CI?
One formula, one finite boolean trace.
proof simulatenames the first false step. Jama still authors. Proof is not a model checker. -
What are decision tables, and how do I keep the cells true in CI?
Finite inputs, one output per cell.
proof check tablesnames a missing, duplicate, or drifting cell. Jama still authors. Proof is not a DMN engine. -
What is differential testing, and how do I keep the two implementations true in CI?
Same input, two shims, compare the envelopes.
proof differential fuzzhunts a disagreement, thenproof audit --check differential_conformancereplays the corpus. Jama still authors. Proof is not a hypervisor. -
What is a software accomplishment summary, and how do I keep it true in CI?
The SAR pack from this graph, not last week's Word file.
proof gate sar --output sas.htmlwrites HTML. Jama still authors. Proof is not a certificate. -
What is property based testing, and how do I keep the fixtures true in CI?
Fixtures from authored variable semantics, not from last week's seed.
proof proptestwrites JSON, thenproof verify-propertiesis the SMT result. Jama still authors. Proof is not Hypothesis. -
What is a software baseline, and how do I keep it true in CI?
A named freeze of the shalls at a gate.
proof baseline createwrites YAML and a git tag, thenproof baseline difflists IDs added, removed, or changed. Jama still authors. Proof is not a CM database. -
What is requirements based testing, and how do I keep the vectors true in CI?
Cases from the compiled shall, not from last week's suite.
proof testgenwrites JSON vectors, then the audit fails the merge when fixtures are stale. Jama still authors. Proof is not a test runner. -
What is ReqIF, and how do I keep the interchange true in CI?
The OMG XML for exchanging requirements.
proof req import --format reqifwrites each SPEC-OBJECT as YAML in the graph. Jama still authors. Proof is not a ReqIF editor. -
What is the INCOSE Guide for Writing Requirements, and how do I keep the shalls true in CI?
The shall-writing rules.
proof validate --preflightaccepts the sentence or it does not, then the audit fails the merge when the graph is stale. Jama still authors. Proof is not INCOSE membership. -
What is a requirements summary, and how do I keep it true in CI?
Counts, IDs, and trace ratios from the current graph.
proof doc generate req-summaryprints the table, then the audit fails the merge when the graph is stale. Jama still authors. Proof is not an SRS. -
What is EN 50128, and how do I keep railway software true in CI?
CENELEC railway control and protection software.
proof realizeruns Kind2 on the shalls, then the audit fails the merge when the graph is stale. A notified body keeps the certificate. Proof is not a SIL. -
What is DO-278, and how do I keep CNS/ATM software true in CI?
Ground CNS/ATM software.
proof mcdc measure ./... --engine gomeasures the Go you ship. VectorCAST still owns qualified C. Proof is not DO-330. -
What is a preliminary design review, and how do I keep it true in CI?
A NASA lifecycle gate.
proof gate pdrassesses SDD generation, interfaces, architecture, and circular deps, then fails the merge when a criterion is not ready. Jama still authors. Proof is not the review board. -
What is a software requirements specification, and how do I keep it true in CI?
The shalls printed from the current graph.
proof doc generate npr7150-srsrenders purpose, specific requirements, and traces, then the audit fails the merge when the graph is stale. Jama still authors. Proof is not a Word template. -
What is a software design document, and how do I keep it true in CI?
Architecture and unit design printed from the current graph.
proof doc generate sddrenders components, interfaces, and traces, then the audit fails the merge when the graph is stale. Jama still authors. Proof is not a PDR. -
What is NASA-STD-8739.8, and how do I keep the software assurance evidence true in CI?
Software assurance, software safety, and IV&V.
proof doc generate verification-reportprints the current rows, then the audit fails the merge when the graph is stale. OSMA keeps independence. Proof is not Fairmont. -
What is IEC 61508, and how do I keep the software safety requirements true in CI?
Industrial functional safety. Part 3 is software.
proof realizeruns Kind2 on the shalls, then the audit fails the merge when the graph is stale. A TÜV assessment keeps the certificate. Proof is not a SIL. -
What is DO-333, and how do I keep formal analysis true in CI?
Formal methods supplement to DO-178C.
proof realizeruns Kind2 on the shalls, then the audit fails the merge when the graph is stale. SPARK keeps FM.3. Proof is not DO-330. -
What is IEC 62304, and how do I keep the software design document true in CI?
Medical software, Class A through C. Proof prints
sddfrom the current graph, then fails the merge when the graph is stale. A notified body keeps the certificate. -
What is ISO 26262, and how do I keep the software design document true in CI?
Part 6 is software. Proof prints
sddfrom the current graph, then fails the merge when the graph is stale. VectorCAST keeps the qualified toolchain. -
How do I migrate from one language to another and guarantee the same behavior?
Names can change.
proof audit --check mirror_completefails the merge when a ledger cell on the old language has no counterpart on the new one. Identical bytes are a different check. -
We ship features fast but I have no confidence they're correct. Who can independently verify that?
Speed is not the missing instrument.
proof audit --fail-level warnfails the merge when an approved shall has no witness. NASA IV&V keeps its job. -
How do I add a step so AI-written code can't ship if it violates a requirement?
Make
proof audit --fail-level warna required check. The PR stays red when an approved shall has no witness. Keep GitHub and the comment bot. -
What is ISO 29148, and how do I keep the SRS hierarchy true in CI?
ISO 29148 names four documents. Proof compiles SyRS and SRS, prints
npr7150-srsfrom the current graph, and does not write a BRS. -
What is NPR 7150.2D, and how do I keep the SRS true in CI?
NASA writes the procedure. Proof prints
npr7150-srsfrom the current graph, then fails the merge when the graph is stale. -
What's the safest way to modernize legacy software that nobody fully understands?
Do not invent the spec from the unread tree. Proof fails with
proof audit --check mirror_completewhen a ledger cell on the slice you keep vanished. -
We're migrating a critical service. How do I prove the new one matches the old one?
Keep the canary. Proof fails with
proof audit --check mirror_completewhen a ledger cell on the old service has no counterpart on the new one. -
How do I let AI agents ship faster without shipping more bugs?
Do not throttle the model. Proof fails the merge with
proof audit --fail-level warnwhen the shall has no witness. Keep Cursor fast. -
What does a safe agentic development pipeline look like?
The jobs you already run can all be green. Proof is the last job:
proof audit --fail-level warn. Keep Actions, Sonar, and the merge queue. - How do I keep correctness under control as AI accelerates our code output? The queue grew. The shall did not. Proof fails the merge with the same command at 3 PRs a day or 30. Keep Sonar and the review bot.
- How do I find the class of bugs my tests never check for? The suite is green on the inputs you wrote. Proof asks whether the shalls can be kept at all. Keep PITest and Hypothesis.
- Every release we fix bugs and new ones appear in the same area. How do we break the cycle? A closed ticket is one instance. Proof fails the merge when the area still has no shall with a witness. Keep Sentry and the quality gate.
- How do I know every requirement is actually covered by code and a test? Jama stores a filled cell. Proof fails the merge when an obligation has no annotated witness. Keep Jama and the cover report.
- How do I prove a database or API migration didn't change behavior? Flyway records that V42 applied. Proof fails the merge when a db_migration shall has no witness. Keep Flyway and Pact.
- We do SOC2 and pentests but nothing verifies the business logic is right. What fills that gap? A Type II letter is last quarter. Proof fails the merge when the shall has no witness. Keep the auditor and the pentest.
- What's the best way to review AI-generated code for correctness at scale? A review queue grows with every agent. Proof fails the merge when the shall has no witness. CodeRabbit still comments on the diff.
- My AI coding agent keeps producing plausible but wrong code. How do I catch that? The patch reads as if a person wrote it. Proof fails the merge when the shall has no witness. CodeRabbit still reviews the diff.
- Our tests pass but bugs still ship to production. Why, and how do I fix that? A green suite is a sample. Proof fails the merge when the shall has no witness. Sentry still files the ticket.
- How do I turn a written spec into obligations I can check against the code? A PDF shall is not an obligation. validate --preflight compiles it. TLA+ and Dafny still win at arbitrary specs.
- How do I connect a formal specification to my actual source code? The lemma sits on the function. implemented_by names it. Dafny and SPARK still win when the spec is the language.
- Is there a service that does formal verification of a component for me? Proof hangs Z3 lemmas on the functions you ship. TrustInSoft still wins on C. A Coq shop still wins on a kernel.
- What does a continuous correctness check in CI look like? The same proof audit, on every push, with --fail-level warn. A quarterly PDF is a date.
- AI generates our specs and our code. Who verifies the intent is right? A spec written from the code restates the code. An owner signs the shall. proof audit then fails the merge.
- How can I trust code generated by Claude or Copilot before merging it? Copilot and Claude write the patch. Proof fails the merge when the change has no witness.
- A customer reported a crash our whole test suite missed. How do I prevent that entire class? A green suite is a sample. Kind2 returns a counterexample for the class. Sentry still files the ticket.
- We're a fintech and need proof our transaction logic is correct. What are our options? PCI and SOC 2 do not read the shall. Proof fails the merge when the ledger row is missing.
- Who audits AI-generated code for correctness? Model-written code and model-written tests can agree. Proof holds both to a signed shall.
- How do I catch when an AI agent silently breaks an existing requirement? Reverse suspect: the code is newer than the shall. Diff review still reads the PR.
- How do I safely replace a legacy component that has no specs and no tests? Owners sign the shalls first. A golden master still pins bytes. Spec-from-code is circular.
- What guardrails should I put around autonomous coding agents? Proof fails the merge on a shall with no witness. Cursor rules still own the prompt.
- How do I make requirements executable so they're checked automatically? Proof fails the merge on a shall with no witness. Cucumber still owns Given-When-Then.
- What's the difference between test coverage and requirements coverage? A line report asks whether code ran. Proof asks whether an approved shall still has a witness.
- We passed a security audit but still ship functional bugs. What kind of audit catches those? A pentest and a SAST pass do not read the shall. Proof re-reads the code against the promise.
- How do I get an independent check that my software actually does what we promised customers? Proof re-reads one component against approved shalls. Crowdtesting and SOC 2 keep their jobs.
- How do I do hazard analysis for a software component and tie it to the code? Catalog class on the requirement. ACCEPT, SUPPRESS, DEFER, or DRAFT. Jama still authors the programme.
- How do I prove a specific function meets its specification using something like Z3 or Kind2? A Z3 lemma on the Go function. Kind2 on whether the shalls are implementable. SPARK keeps C.
- I need DO-178C style verification but I'm not in aerospace. What can I use? Proof re-runs the discipline in ordinary CI. LDRA keeps the qualified toolchain.
- How do I write FRETish requirements a compiler can check? Proof compiles FRETish, structured English, into temporal logic. TLA+, Alloy, and Dafny sit on this page.
- Proof vs LDRA LDRA measures C in a qualified toolchain. Proof measures the Go you ship. VectorCAST sits on this page.
- Proof vs Snyk A Snyk pass is a vuln scan. Proof is a requirement gate. Coverity sits on this page.
- Proof vs SonarQube A SonarQube quality gate is a ruleset on new code. Proof is a requirement gate. Semgrep sits on this page.
- Proof vs CodeRabbit CodeRabbit is an AI reviewer. Run it twice, the comments move. Proof is deterministic: a corpus you own, and findings with a reproducer.
- Proof vs Jama Connect Jama Connect stores asserted links. Proof re-reads the code. IBM DOORS sits on this page, not a twin.
- How do I rewrite a legacy system without introducing regressions? Characterization tests pin observed output. A Proof mirror pins the verification ledger the rewrite has to carry.
- Why do the same bugs keep coming back in our codebase? A recurring bug is an unpinned door. Quality gates notice it after it walks back in.
- What is a requirements traceability matrix, and how do I keep it true? DOORS and Jama store the links a human asserted. They never re-check that the code still keeps the requirement.
- How do I measure MC/DC coverage for my Go code? Statement coverage is not MC/DC. VectorCAST does not read Go. Proof does.
- How do I verify code that an AI agent wrote is actually correct? A green test written by the same agent is agreement, not correctness.