Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
import { readFileSync } from 'fs';
import { describe, expect, it } from 'vitest';
import { repoRoot, runPromptMigrationTests } from './_testHelpers.js';
import migration, {
ACCEPTED_OLD_MD5,
NEW_SHIPPED_MD5,
applyMigration,
} from './368-fableloom-plan-review-evolution.js';
// Wrap-insensitive: the template is hard-wrapped at 80 columns, so a phrase
// assertion keyed to today's line breaks would fail on a pure re-wrap that
// changed no guidance at all.
const body = () => readFileSync(
`${repoRoot}/data.reference/prompts/stages/fableloom-review-series-plan.md`,
'utf8',
).replace(/\s+/g, ' ');
describe('migration 368 — FableLoom plan review evolution lens', () => {
runPromptMigrationTests({
migration,
applyMigration,
ACCEPTED_OLD_MD5,
NEW_SHIPPED_MD5,
prefix: 'migration-368-fableloom-plan-review-evolution-',
});
it('gates the lens behind the variable weave.js supplies', () => {
// Without the mustache SECTION the block renders even with no lens authored
// — which breaks the "unchanged review for a plan with no lens" promise.
// Without the VARIABLE the section renders empty forever.
const md = body();
expect(md).toContain('{{#characterEvolutions}}');
expect(md).toContain('{{characterEvolutions}}');
expect(md).toContain('{{/characterEvolutions}}');
});
it('states the causal chain and rules every declared outcome', () => {
const md = body();
// The chain is what the plan digest alone cannot express; shipping the lens
// data without it leaves the model to guess the mapping.
for (const stage of [
'control strategy failing',
'pressure forces exploration',
'commitment to change',
'cost tested',
'final proof',
]) {
expect(md).toContain(stage);
}
// An unruled outcome falls back to the pre-lens reading, which is the bug:
// a tragic refusal or a deliberate flat arc reported as an arc gap.
for (const outcome of ['full-change', 'tragic-refusal', 'flat-testing', 'partial-open', 'undeclared']) {
expect(md).toContain(`\`${outcome}\``);
}
expect(md).toContain('do not report it as a flat arc');
expect(md).toContain('UNEARNED CLAIM OF COMPLETION');
});
it('makes an external victory with no behavioral proof a risk, not a strength', () => {
// The single behavior #6418 names as the reason the lens reaches the plan
// review at all.
const md = body();
expect(md).toContain('not the defeat of an external obstacle');
expect(md).toContain('is a risk, not a strength');
});
it('keeps the two failure modes separable and demands a concrete repair', () => {
const md = body();
// They need different repairs, so the risk text must stay separable by a
// reader (and by a downstream filter).
expect(md).toContain('`No authored intent:`');
expect(md).toContain('`Authored intent not delivered:`');
expect(md).toContain('must name a concrete scene or plan repair');
// A stale/unverified anchor can never read as verified (epic non-negotiable).
expect(md).toContain('[stale]');
expect(md).toContain('[unverified]');
expect(md).toContain('NOT PROVEN');
});
it('lets a satisfied lens end the planning loop', () => {
// `editorialAutopilot.runPlanning()` re-runs feedback while `risks` is
// non-empty. A prompt that always emits an evolution risk turns that into a
// maxRounds spin, so the template has to say the empty answer is correct.
const md = body();
expect(md).toContain('this pass contributes nothing');
expect(md).toContain('An empty `risks` array is the correct answer');
// No page counts / episode quotas / literal-opposite belief.
expect(md).toContain('NOT a page count, a chapter count, an episode quota');
});
});