Spot it. Save it. Free, local-first Chrome extension for images, videos and prompts. Explore 2,000+ curated cases and connect your library to AI Agents via MCP. 提示词导演
JavaScript ai-artai-videochrome-extensionclaude-codecodexhiggsfieldjimenglocal-firstmcpmidjourneymoodboardprompt-engineeringprompt-libraryprompt-managerreference-libraryseedancevisual-bookmarksweb-clipper
import test from 'node:test';
import assert from 'node:assert/strict';
import { detailPromptSources } from '../extension/prompt-sources.js';
import { canonicalTextAnalysisInput } from '../extension/analysis-input.js';
import { caseTextPart } from '../extension/composer-library-tools.js';
import { composerSourceText } from '../extension/composer-source-text.js';
import { createAgentLibrary } from '../extension/agent-library.js';
import { classifyImportedMedia } from '../extension/classifier.js';
import { CONTENT_IDS } from '../extension/taxonomy.js';
import { renderLibraryJson } from '../extension/lib.js';
function fixture(kind) {
return { id:`case-${kind}`, title:'原始资料', text:'来源中的完整提示词', textRevision:4,
sourceFacts:{originalPromptAvailable:true},
classification:{pathIds:[kind==='video'?CONTENT_IDS.videoCase:CONTENT_IDS.imageCase],status:'confirmed',source:'manual'},
mediaAssets:[{id:'asset',kind,usage:'content'}], primaryMediaId:'asset' };
}
for (const kind of ['image','video']) test(`${kind}: original remains readable across detail, text analysis, composer, Agent and export despite user classification`, async () => {
const entry = fixture(kind), before = structuredClone(entry);
const library = createAgentLibrary({loadState:async()=>({entries:[entry]}),libraryUrl:'chrome-extension://fixture/library.html'});
assert.equal(detailPromptSources(entry,entry.mediaAssets[0]).original,entry.text);
assert.equal(canonicalTextAnalysisInput(entry).text,entry.text);
assert.equal(caseTextPart(entry,'original_prompt'),entry.text);
assert.equal(composerSourceText(entry),entry.text);
assert.equal((await library.read({caseId:entry.id,part:'original_prompt'})).content,entry.text);
assert.equal(JSON.parse(renderLibraryJson([entry],{})).entries[0].text,entry.text);
assert.deepEqual(entry,before);
});
test('video analysis uses its selected original, not a poster or adopted AI text', () => {
const entry=fixture('video');
entry.mediaAssets.unshift({id:'poster',kind:'image',usage:'poster'});
entry.mediaPrompts=[{assetId:'asset',text:'AI推测',source:'ai-suggestion'},
{assetId:'asset',text:'视频原始输入',source:'manual',textRevision:6}];
assert.deepEqual(canonicalTextAnalysisInput(entry),{text:'视频原始输入',textRevision:6,source:'media_prompt',assetId:'asset'});
assert.equal(detailPromptSources(entry,entry.mediaAssets[1]).original,'视频原始输入');
assert.doesNotMatch(caseTextPart(entry,'original_prompt'),/AI推测/);
});
test('an adopted AI prompt remains analyzable when no original is present, without becoming an original', () => {
const entry=fixture('image');
entry.text='';
entry.sourceFacts.originalPromptAvailable=false;
entry.mediaPrompts=[{assetId:'asset',text:'用户已采用的AI提示词',source:'ai-suggestion',textRevision:2}];
assert.deepEqual(canonicalTextAnalysisInput(entry),{text:'用户已采用的AI提示词',textRevision:2,source:'ai_prompt',assetId:'asset'});
assert.equal(detailPromptSources(entry,entry.mediaAssets[0]).original,'');
assert.equal(caseTextPart(entry,'original_prompt'),'');
});
test('article and compound reading never promotes article prose to original prompts', () => {
const article={id:'article',title:'教程',text:'文章正文',articleDocument:{blocks:[{kind:'paragraph',text:'文章正文'}]}};
const video=fixture('video');
const compound={id:'compound',text:'文章正文\n来源中的完整提示词',memberEntries:[article,video]};
assert.equal(caseTextPart(article,'original_prompt'),'');
assert.match(caseTextPart(compound,'original_prompt'),/来源中的完整提示词/);
assert.doesNotMatch(caseTextPart(compound,'original_prompt'),/文章正文/);
});
test('new file imports classify from recorded original evidence while respecting manual organization', () => {
for (const kind of ['image','video']) {
const entry=fixture(kind);
const manual=entry.classification;
assert.deepEqual(classifyImportedMedia(entry).pathIds,manual.pathIds);
assert.equal(classifyImportedMedia(entry).source,manual.source);
delete entry.classification;
assert.deepEqual(classifyImportedMedia(entry).pathIds,[kind==='video'?CONTENT_IDS.promptVideo:CONTENT_IDS.promptImage]);
entry.sourceFacts={originalPromptAvailable:false};entry.text='随手备注';
assert.deepEqual(classifyImportedMedia(entry).pathIds,[kind==='video'?CONTENT_IDS.videoCase:CONTENT_IDS.imageCase]);
entry.mediaPrompts=[{assetId:'asset',source:'embedded',text:'文件中真实原词'}];
assert.deepEqual(classifyImportedMedia(entry).pathIds,[kind==='video'?CONTENT_IDS.promptVideo:CONTENT_IDS.promptImage]);
}
});