wzyn20051216/matlab-agent-skills / skills/matlab-testing-ci
matlab-testing-ci
VerifiedMATLAB testing, validation, and CI workflow for matlab.unittest, MATLAB Test, Simulink Test, coverage, artifact generation, GitHub Actions, and open-source reproducibility. Use whenever MATLAB work needs verification, regression tests, CI, GitHub publication, or acceptance criteria.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.5 KB |
| assets/install-roots.json | config | 2.9 KB |
| scripts/mas_bench_record.m | other | 11.5 KB |
| scripts/mas_capability_check.m | other | 12.7 KB |
| scripts/mas_delivery_check.m | other | 32.2 KB |
| scripts/mas_grade.m | other | 5.5 KB |
| scripts/mas_phase43_grade.m | other | 12.4 KB |
| scripts/mas_ref_derive.m | other | 21.7 KB |
| scripts/mas_ref_spec.m | other | 6.3 KB |
| scripts/mas_refcheck.m | other | 10.7 KB |
| scripts/mas_skill_audit.m | other | 10.9 KB |
| scripts/mas_skill_install_audit.m | other | 13.7 KB |
| scripts/mas_verify_task.m | other | 3.9 KB |
| tests/test_bench.m | other | 5.4 KB |
| tests/test_capability.m | other | 6.1 KB |
| tests/test_delivery_check.m | other | 22.0 KB |
| tests/test_grade.m | other | 3.0 KB |
| tests/test_phase43_grade.m | other | 9.1 KB |
| tests/test_refcheck.m | other | 5.0 KB |
| tests/test_skill_install_audit.m | other | 8.7 KB |
tests/test_delivery_check.m
View on GitHubfunction res = test_delivery_check(verbose)
%TEST_DELIVERY_CHECK Format and engineering math must fail independently.
%
% res = test_delivery_check()
%
% The cases below are chosen so that every combination that was previously
% collapsed into one boolean appears at least once:
% * right numbers, wrong schema -> format FAIL, math PASS
% * right schema, wrong numbers -> format PASS, math FAIL
% * wrong verdict with a valid schema -> verdict FAIL, never coerced
% * legal UNKNOWN / UNDETERMINED -> not a schema error
% * missing field / non-finite value -> UNDETERMINED, not a pass
% * legacy result without units -> units UNRECORDED, not "not required"
% * tampered reference manifest -> reference FAIL, math untouched
if nargin < 1, verbose = true; end
here = fileparts(mfilename('fullpath'));
addpath(fullfile(fileparts(here), 'scripts'));
tmp = fullfile(tempdir, sprintf('mas_delivery_%s', datestr(now, 'yyyymmddHHMMSSFFF')));
if ~isfolder(tmp), mkdir(tmp); end
cleaner = onCleanup(@() cleanup_dir(tmp)); %#ok<NASGU>
spec = struct();
spec.schema = 'mas-delivery-contract/1';
spec.profile = 'mas-result/2';
spec.taskId = 'P57-DLV-1';
spec.expectedVerdict = 'PASS';
spec.fields = struct( ...
'name', {'settlingTime', 'overshoot', 'mode'}, ...
'op', {'abs', 'abs', 'exact'}, ...
'ref', {4.0, 16.3, 'stable'}, ...
'tol', {0.05, 0.2, 0}, ...
'unit', {'s', '%', ''});
spec.requiredFields = {'settlingTime', 'overshoot', 'mode'};
spec.units = struct('settlingTime', 's', 'overshoot', '%');
spec.enums = struct('verdict', {{'PASS', 'FAIL', 'UNDETERMINED', 'UNKNOWN'}});
spec.budget = struct('required', true, 'maxToolCalls', 8, 'maxWallSeconds', 120);
r = struct();
r.schema = 'mas-result/2';
r.taskId = 'P57-DLV-1';
r.verdict = 'PASS';
r.fields = struct('settlingTime', 4.02, 'overshoot', 16.25, 'mode', 'stable');
r.units = struct('settlingTime', 's', 'overshoot', '%');
r.budget = struct('toolCalls', 5, 'wallSeconds', 41);
nP = 0; nT = 0; fails = {};
if verbose, fprintf('=== delivery contract tests ===\n'); end
% ---- 1. a fully valid delivery --------------------------------------------
A = run_check(r, spec, {}, false);
chk_has(A.schema, 'mas-delivery-check/1', 'output schema');
chk(A.overall, 'PASS', 'valid delivery passes');
chk(A.format.status, 'PASS', 'format accepted');
chk(A.required.status, 'PASS', 'required fields present');
chk(A.units.status, 'PASS', 'units match');
chk(A.verdict.status, 'PASS', 'verdict matches');
chk(A.math.status, 'PASS', 'math within tolerance');
chk(A.budget.status, 'PASS', 'budget honoured');
chk(A.reference.status, 'NOT_REQUIRED', 'reference not requested');
chk(A.readOnly, true, 'checker declares itself read-only');
chk(A.verdictCoerced, false, 'checker never coerces');
chk(isempty(A.errors), true, 'no errors on a clean delivery');
chk(~A.separation.numericValidFormatInvalid && ~A.separation.formatValidNumericInvalid, ...
true, 'no separation hint when both pass');
chk(A.math.engine, 'mas_grade', 'math delegates to mas_grade');
chk(isfield(A.math.grade, 'correct') && isfield(A.math.grade, 'meets'), true, ...
'mas_grade keeps ran/correct/meets separate');
% ---- 2. wrong verdict: FAIL against a FAIL expectation ---------------------
r2 = r; r2.verdict = 'FAIL';
f2 = fullfile(tmp, 'result_wrong_verdict.json');
w(f2, jsonencode(r2));
raw2 = fileread(f2);
rep2 = fullfile(tmp, 'report_wrong_verdict.json');
A = run_check(f2, spec, {'outFile', rep2}, false);
chk(A.verdict.status, 'FAIL', 'wrong verdict fails the verdict dimension');
chk(A.verdict.actual, 'FAIL', 'actual verdict is reported as delivered');
chk(A.verdict.expected, 'PASS', 'expected verdict is reported');
chk(has_error(A, 'VERDICT_WRONG'), true, 'wrong verdict is a structured error');
chk(A.math.status, 'PASS', 'math still passes: numbers are not the problem');
% A wrong verdict is neither a format error nor a math error. The separation
% flags cover exactly those two axes, so both stay false and the verdict has
% its own dimension and its own error code.
chk(A.separation.numericValidFormatInvalid, false, ...
'verdict failure is not mislabelled as a format error');
chk(A.separation.formatValidNumericInvalid, false, ...
'verdict failure is not mislabelled as a math error');
chk(strcmp(fileread(f2), raw2), true, 'result file is never rewritten');
chk(A.verdictCoerced, false, 'wrong verdict is not rewritten to match');
P = jsondecode(fileread(rep2));
chk(P.schema, 'mas-delivery-check/1', 'report file carries the output schema');
% ---- 3. correct numbers, invalid schema -----------------------------------
r3 = rmfield(r, {'schema', 'taskId', 'units'});
A = run_check(r3, spec, {}, false);
chk(A.format.status, 'FAIL', 'missing schema/taskId fails the format dimension');
chk(has_error(A, 'SCHEMA_VERSION'), true, 'schema version error recorded');
chk(has_error(A, 'TASK_ID_MISSING'), true, 'task id error recorded');
chk(A.units.status, 'FAIL', 'missing units map fails the units dimension');
chk(has_error(A, 'UNITS_MISSING'), true, 'units error recorded');
chk(A.math.status, 'PASS', 'math passes although the schema does not');
chk(A.separation.numericValidFormatInvalid, true, 'hint separates math from format');
chk_has(A.separation.hint, 'fix the format, not the number', 'format hint text');
chk(A.overall, 'FAIL', 'invalid schema fails the delivery overall');
% ---- 4. a required field is absent ---------------------------------------
r4 = r; r4.fields = rmfield(r.fields, 'overshoot');
A = run_check(r4, spec, {}, false);
chk(A.required.status, 'FAIL', 'missing required field fails');
chk(has_error(A, 'REQUIRED_FIELD_MISSING'), true, 'structured missing-field error');
chk(A.math.status, 'UNDETERMINED', 'absent field leaves math undetermined');
chk(A.overall, 'FAIL', 'missing required field fails overall');
% ---- 5. UNKNOWN / UNDETERMINED are legal states --------------------------
spec5 = spec; spec5.expectedVerdict = 'UNDETERMINED';
spec5.fields(3).ref = 'UNKNOWN';
r5 = r; r5.verdict = 'UNDETERMINED'; r5.fields.mode = 'UNKNOWN';
A = run_check(r5, spec5, {}, false);
chk(A.verdict.status, 'PASS', 'UNDETERMINED matches an UNDETERMINED expectation');
chk(A.math.status, 'PASS', 'UNKNOWN matches an UNKNOWN reference exactly');
chk(A.enums.status, 'PASS', 'legal states are not enum errors');
chk(A.overall, 'PASS', 'legal unknown states can still pass');
r5b = r5; r5b.verdict = 'UNKNOWN';
A = run_check(r5b, spec5, {}, false);
chk(A.verdict.status, 'FAIL', 'a legal but different verdict still fails');
chk(A.enums.status, 'PASS', 'that failure is not a schema error');
% ---- 6. an illegal enum value -------------------------------------------
r6 = r; r6.verdict = 'MAYBE';
A = run_check(r6, spec, {}, false);
chk(A.enums.status, 'FAIL', 'illegal enum value fails the enum dimension');
chk(has_error(A, 'ENUM_VALUE'), true, 'enum error recorded');
chk(A.verdict.status, 'FAIL', 'illegal verdict fails the verdict dimension');
chk(has_error(A, 'VERDICT_ILLEGAL'), true, 'illegal verdict recorded');
chk(A.overall, 'FAIL', 'illegal enum fails overall');
% ---- 7. units ------------------------------------------------------------
r7 = r; r7.units.settlingTime = 'ms';
A = run_check(r7, spec, {}, false);
chk(A.units.status, 'FAIL', 'unit mismatch fails the units dimension');
chk(has_error(A, 'UNIT_MISMATCH'), true, 'unit mismatch recorded');
chk(A.math.status, 'PASS', 'unit mismatch does not invent a math failure');
chk(A.overall, 'FAIL', 'unit mismatch fails overall');
r7b = r; r7b.units.mass = 'kg';
A = run_check(r7b, spec, {}, false);
chk(A.units.status, 'PASS', 'unknown extra unit is a warning, not a failure');
chk(has_error(A, 'UNIT_UNKNOWN'), true, 'unknown unit warning recorded');
chk(A.overall, 'PASS', 'warning does not fail the delivery');
% ---- 8. right schema, wrong number --------------------------------------
r8 = r; r8.fields.settlingTime = 4.5;
A = run_check(r8, spec, {}, false);
chk(A.format.status, 'PASS', 'schema stays valid');
chk(A.math.status, 'FAIL', 'out-of-tolerance number fails math');
chk(has_error(A, 'MATH_MISMATCH'), true, 'math mismatch recorded');
chk(A.separation.formatValidNumericInvalid, true, ...
'separation flags schema-valid / numbers-invalid');
chk_has(A.separation.hint, 'still wrong', 'math hint text');
chk(A.overall, 'FAIL', 'wrong number fails overall');
% ---- 9. non-finite evidence is not a pass -------------------------------
r9 = r; r9.fields.overshoot = NaN;
A = run_check(r9, spec, {}, false);
chk(A.math.status, 'UNDETERMINED', 'NaN leaves math undetermined');
chk(A.overall, 'UNDETERMINED', 'undetermined math makes the delivery undetermined');
% ---- 10. budget ---------------------------------------------------------
r10 = r; r10.budget.toolCalls = 12;
A = run_check(r10, spec, {}, false);
chk(A.budget.status, 'FAIL', 'budget overflow fails');
chk(has_error(A, 'BUDGET_EXCEEDED'), true, 'budget overflow recorded');
chk(A.overall, 'FAIL', 'budget overflow fails overall');
r10b = rmfield(r, 'budget');
A = run_check(r10b, spec, {}, false);
chk(A.budget.status, 'UNRECORDED', 'unrecorded usage is not a pass');
chk(A.overall, 'UNDETERMINED', 'unrecorded budget makes the delivery undetermined');
% ---- 11. unreadable / missing results -----------------------------------
f11 = fullfile(tmp, 'broken.json');
w(f11, '{ this is not json');
A = run_check(f11, spec, {}, false);
chk(A.overall, 'ERROR', 'unreadable result is an ERROR');
chk(has_error(A, 'RESULT_UNREADABLE'), true, 'unreadable result recorded');
A = run_check(fullfile(tmp, 'does_not_exist.json'), spec, {}, false);
chk(A.overall, 'ERROR', 'missing result is an ERROR');
chk(has_error(A, 'RESULT_MISSING'), true, 'missing result recorded');
% ---- 12. legacy profile: honest about what it cannot prove --------------
spec12 = spec; spec12.profile = 'mas-result/1-legacy';
r12 = struct('verdict', 'PASS', 'fields', ...
struct('settlingTime', 4.0, 'overshoot', 16.3, 'mode', 'stable'));
A = run_check(r12, spec12, {}, false);
chk(A.format.status, 'PASS', 'legacy shape passes the legacy format rules');
chk(A.format.schemaVersion, 'legacy/none', 'legacy shape is labelled');
chk(A.units.status, 'UNRECORDED', 'legacy result cannot prove unit consistency');
chk(A.overall, 'UNDETERMINED', 'legacy delivery is honest about the gap');
% ---- 13. unknown profile -------------------------------------------------
spec13 = spec; spec13.profile = 'mas-result/9';
A = run_check(r, spec13, {}, false);
chk(A.format.status, 'FAIL', 'unknown profile fails the format dimension');
chk(has_error(A, 'PROFILE_UNKNOWN'), true, 'unknown profile recorded');
% ---- 14. task identity ---------------------------------------------------
r14 = r; r14.taskId = 'P57-DLV-OTHER';
A = run_check(r14, spec, {}, false);
chk(A.format.status, 'FAIL', 'wrong task id fails the format dimension');
chk(has_error(A, 'TASK_ID_MISMATCH'), true, 'task id mismatch recorded');
% ---- 15. exact string mismatch ------------------------------------------
r15 = r; r15.fields.mode = 'unstable';
A = run_check(r15, spec, {}, false);
chk(A.math.status, 'FAIL', 'exact string mismatch fails math');
chk(has_error(A, 'MATH_MISMATCH'), true, 'exact mismatch recorded');
chk(A.overall, 'FAIL', 'exact mismatch fails overall');
% ---- 16. a tampered reference manifest ----------------------------------
% Only the D3 entry is copied, so verification re-derives one reference
% instead of fifteen. The hash is zeroed: values still match the derivation,
% which isolates the hash guard from the value guard.
repo = fileparts(fileparts(fileparts(here)));
realManifest = fullfile(repo, 'evals', 'references', 'reference_manifest.json');
RM = jsondecode(fileread(realManifest));
d3 = RM.entries.D3;
d3.hash = repmat('0', 1, numel(char(d3.hash)));
fake = struct('schema', 'mas-reference/1', 'entries', struct('D3', d3));
f16 = fullfile(tmp, 'tampered_manifest.json');
w(f16, jsonencode(fake));
A = run_check(r, spec, {'referenceManifest', f16}, false);
chk(A.reference.status, 'FAIL', 'tampered manifest fails the reference dimension');
chk(has_error(A, 'REFERENCE_MISMATCH'), true, 'reference mismatch recorded');
chk(A.math.status, 'PASS', 'reference failure does not contaminate math');
chk(A.overall, 'FAIL', 'tampered reference fails overall');
% ---- 17. the versioned schema file agrees with the checker --------------
schemaFile = fullfile(repo, 'schemas', 'mas-result-2.schema.json');
chk(isfile(schemaFile), true, 'versioned result schema exists');
J = jsondecode(fileread(schemaFile));
chk(J.properties.schema.const, 'mas-result/2', 'schema pins the version string');
chk(isequal(sort(J.properties.verdict.enum(:))', ...
sort({'PASS', 'FAIL', 'UNDETERMINED', 'UNKNOWN'})), true, ...
'schema enum matches the legal verdict states');
req = J.required(:)';
chk(all(ismember({'schema', 'taskId', 'verdict', 'fields'}, req)), true, ...
'schema requires the fields the checker enforces');
chk(J.properties.fields.type, 'object', 'schema requires fields to be an object');
% ---- 18..27 v3 dimensions: seven axes reported independently ------------
% A vocabulary mistake must not be counted as a numerical mistake, and a
% numerically correct answer must not rescue a broken delivery contract.
specW = spec;
specW.workflow = struct('required', true, ...
'stages', {{'skillSelected', 'referencesLoaded', 'mcpExecuted', 'checksRun'}});
rW = r;
rW.workflow = struct('skillSelected', true, 'referencesLoaded', true, ...
'mcpExecuted', true, 'checksRun', true);
A = run_check(rW, specW, {}, false);
chk(A.dimensionSchema, 'mas-delivery-dimensions/3', 'v3 dimension schema is labelled');
chk(A.schema, 'mas-delivery-check/1', 'v1 output schema is unchanged');
d = A.dimensions;
chk(d.numericalCorrectness.status, 'PASS', 'v3 numerical axis passes on a clean delivery');
chk(d.enumContractValid.status, 'PASS', 'v3 enum axis passes on a clean delivery');
chk(d.engineeringVerdictCorrect.status, 'PASS', 'v3 verdict axis passes');
chk(d.structuralSchemaValid.status, 'PASS', 'v3 structural axis passes');
chk(d.workflowCompleted.status, 'PASS', 'v3 workflow axis passes');
chk(d.withinBudget.status, 'PASS', 'v3 budget axis passes');
chk(d.deliveryValid.status, 'PASS', 'v3 delivery is valid');
chk(numel(fieldnames(d)), 7, 'exactly seven v3 dimensions');
chk(numel(A.mapping.dimensions) == 7 && ...
strcmp(A.mapping.dimensions(7).name, 'deliveryValid'), true, ...
'mapping documents all seven dimensions');
chk(A.mapping.enumErrorsCountAsNumericalErrors, false, ...
'mapping states enum errors are not numerical errors');
chk(has_string(A.mapping.dimensions(1).from, 'math.numericStatus'), true, ...
'numericalCorrectness is sourced from math.numericStatus');
chk(strcmp(d.deliveryValid.sources(end), 'reference.status'), true, ...
'deliveryValid also carries the reference guard');
% enum-only failure: right numbers, wrong categorical spelling
r18 = r; r18.fields.mode = 'unstable'; r18.workflow = rW.workflow;
A = run_check(r18, specW, {}, false);
chk(A.math.status, 'FAIL', 'v1 math label stays combined for a wrong enum');
chk(A.dimensions.numericalCorrectness.status, 'PASS', ...
'a categorical spelling error is not a numerical error');
chk(A.dimensions.enumContractValid.status, 'FAIL', ...
'the categorical error lands on the enum dimension');
chk(A.dimensions.engineeringVerdictCorrect.status, 'PASS', ...
'the engineering verdict is still the expected one');
chk(A.dimensions.deliveryValid.status, 'FAIL', ...
'an enum error still fails the delivery contract');
chk(A.overall, 'FAIL', 'v1 overall is unchanged by v3');
% numeric-only failure: right vocabulary, wrong number
r19 = r; r19.fields.settlingTime = 4.5; r19.workflow = rW.workflow;
A = run_check(r19, specW, {}, false);
chk(A.dimensions.numericalCorrectness.status, 'FAIL', ...
'an out-of-tolerance number fails the numerical axis');
chk(A.dimensions.enumContractValid.status, 'PASS', ...
'a wrong number is not an enum error');
chk(A.dimensions.structuralSchemaValid.status, 'PASS', ...
'a wrong number does not break the schema');
chk(A.dimensions.deliveryValid.status, 'FAIL', 'a wrong number fails the delivery');
% illegal verdict: enum axis and verdict axis both fail, math stays clean
r20 = r; r20.verdict = 'MAYBE'; r20.workflow = rW.workflow;
A = run_check(r20, specW, {}, false);
chk(A.dimensions.enumContractValid.status, 'FAIL', 'illegal enum value fails the enum axis');
chk(A.dimensions.engineeringVerdictCorrect.status, 'FAIL', ...
'an illegal verdict fails the verdict axis');
chk(A.dimensions.numericalCorrectness.status, 'PASS', ...
'an illegal verdict is not a numerical error');
% legal but wrong verdict: only the verdict axis fails
r21 = r; r21.verdict = 'FAIL'; r21.workflow = rW.workflow;
A = run_check(r21, specW, {}, false);
chk(A.dimensions.enumContractValid.status, 'PASS', 'FAIL is a legal enum value');
chk(A.dimensions.engineeringVerdictCorrect.status, 'FAIL', ...
'a legal but wrong verdict fails only the verdict axis');
chk(A.dimensions.deliveryValid.status, 'FAIL', 'a wrong verdict fails the delivery');
% structural failure with correct numbers: schema, not math
r22 = rmfield(r, {'schema', 'taskId', 'units'}); r22.workflow = rW.workflow;
A = run_check(r22, specW, {}, false);
chk(A.dimensions.structuralSchemaValid.status, 'FAIL', 'missing schema fails the structural axis');
chk(A.dimensions.numericalCorrectness.status, 'PASS', ...
'the numbers are still correct, so the numerical axis passes');
chk(A.dimensions.enumContractValid.status, 'PASS', 'the enum axis is untouched');
chk(A.separation.numericValidFormatInvalid, true, 'v1 separation flag is preserved');
% workflow axis
A = run_check(r, specW, {}, false);
chk(A.dimensions.workflowCompleted.status, 'UNRECORDED', ...
'a required workflow that is not recorded is not a pass');
chk(A.dimensions.deliveryValid.status, 'UNDETERMINED', ...
'an unrecorded workflow leaves the delivery undetermined');
chk(has_error(A, 'WORKFLOW_UNRECORDED'), true, 'unrecorded workflow is a structured error');
r23 = rW; r23.workflow.mcpExecuted = false;
A = run_check(r23, specW, {}, false);
chk(A.dimensions.workflowCompleted.status, 'FAIL', 'a stage recorded as incomplete fails');
chk(has_error(A, 'WORKFLOW_STAGE_INCOMPLETE'), true, 'incomplete stage is recorded');
chk(A.dimensions.numericalCorrectness.status, 'PASS', ...
'an unfinished workflow is not counted as a numerical error');
r24 = rmfield(rW, 'workflow');
A = run_check(r24, spec, {}, false);
chk(A.dimensions.workflowCompleted.status, 'NOT_REQUIRED', ...
'a contract without workflow does not invent a workflow dimension');
chk(A.dimensions.deliveryValid.status, 'PASS', ...
'NOT_REQUIRED dimensions do not block a valid delivery');
% budget axis
r25 = rW; r25.budget.toolCalls = 12;
A = run_check(r25, specW, {}, false);
chk(A.dimensions.withinBudget.status, 'FAIL', 'budget overflow lands on the budget axis');
chk(A.dimensions.numericalCorrectness.status, 'PASS', 'budget overflow is not a numerical error');
chk(A.dimensions.deliveryValid.status, 'FAIL', 'budget overflow fails the delivery');
% legacy replay keeps its honest gap in both protocols
spec26 = specW; spec26.profile = 'mas-result/1-legacy';
r26 = struct('verdict', 'PASS', 'fields', struct('settlingTime', 4.02, ...
'overshoot', 16.25, 'mode', 'stable'), 'workflow', struct('skillSelected', true, ...
'referencesLoaded', true, 'mcpExecuted', true, 'checksRun', true));
A = run_check(r26, spec26, {}, false);
chk(A.overall, 'UNDETERMINED', 'v1 legacy replay stays undetermined');
chk(A.dimensions.structuralSchemaValid.status, 'UNDETERMINED', ...
'v3 records the unit gap instead of passing it');
chk(A.dimensions.numericalCorrectness.status, 'PASS', ...
'the legacy numbers are still judged on their own axis');
chk(A.dimensions.deliveryValid.status, 'UNDETERMINED', ...
'v3 delivery is undetermined exactly where v1 was');
% error paths still report dimensions
A = run_check(fullfile(tmp, 'does_not_exist.json'), specW, {}, false);
chk(A.overall, 'ERROR', 'missing result is an ERROR in v1');
chk(A.dimensions.deliveryValid.status, 'NOT_RUN', 'v3 dimensions say they did not run');
chk(A.dimensions.numericalCorrectness.status, 'NOT_RUN', ...
'no dimension is silently reported as passing after an ERROR');
res = struct('nPass', nP, 'nTotal', nT, 'pass', nP == nT, 'failures', {fails});
if verbose
fprintf('\n---- %d / %d delivery contract cases passed ----\n', nP, nT);
for k = 1:numel(fails)
fprintf(' FAILED: %s\n', fails{k});
end
end
if ~isempty(fails)
error('test_delivery_check:case', '%d/%d cases failed: %s', ...
numel(fails), nT, strjoin(fails, '; '));
end
function A = run_check(result, sp, opts, v)
A = mas_delivery_check(result, sp, opts{:}, 'verbose', v);
end
function chk(got, want, label)
nT = nT + 1;
if ischar(want)
ok = ischar(got) && strcmp(got, want);
else
ok = isequal(got, want);
end
nP = nP + ok;
if verbose
fprintf(' [%s] %s\n', tern(ok, 'PASS', 'FAIL'), label);
end
if ~ok
fails{end+1} = sprintf('%s (got %s, want %s)', label, tostr(got), tostr(want));
end
end
function chk_has(hay, needle, label)
nT = nT + 1;
ok = ischar(hay) && ~isempty(strfind(hay, needle)); %#ok<STREMP>
nP = nP + ok;
if verbose
fprintf(' [%s] %s\n', tern(ok, 'PASS', 'FAIL'), label);
end
if ~ok
fails{end+1} = sprintf('%s (got "%s", want substring "%s")', ...
label, hay, needle);
end
end
end
% ===========================================================================
function ok = has_error(R, code)
ok = any(strcmp({R.errors.code}, code));
end
function ok = has_string(hay, needle)
ok = ischar(hay) && ~isempty(strfind(hay, needle)); %#ok<STREMP>
end
function w(p, s)
d = fileparts(p);
if ~isfolder(d), mkdir(d); end
fid = fopen(p, 'w');
if fid < 0, error('test_delivery_check:write', 'cannot write %s', p); end
fwrite(fid, s, 'char'); fclose(fid);
end
function s = tern(c, a, b)
if c, s = a; else, s = b; end
end
function s = tostr(v)
if ischar(v), s = v;
elseif islogical(v) && isscalar(v), s = mat2str(v);
elseif isnumeric(v) && isscalar(v), s = num2str(v);
else, s = class(v);
end
end
function cleanup_dir(p)
if isfolder(p), rmdir(p, 's'); end
end
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |