Files
AgentCompany/tests/unit/builtin-skills.test.ts
T
ZacharyZhang-NY 86aa6fdb29 Ship skills built-in: bundle the catalogue, assign per hire, force humanizer
Skills no longer come from the user's ~/.agents library - most users have
none, and the CEO branch that compiled the entire catalogue into its
prompt overflowed the model context (1.1MB) and failed every CEO run.

- Bundle "built-in skills/" with the app (extraResources) and stage it
  at startup into <profile>/builtin-skills as one dir-per-skill copy -
  both bundle shapes (SKILL.md dirs and flat category .md files)
  normalize into the same importable layout
- Import the staged catalogue per org at startup and inside onboarding,
  so the CEO's first run already carries it
- Forced house style: resolveRuntimeSnapshot unions the humanizer skill
  into every agent's runtime (CEO included) as a resolution rule - no
  binding row to forget, delete, or bypass; revoke_skill refuses it
- CEO staffs hires with skills: hire_agent accepts a skills array,
  grant_skill/revoke_skill adjust later, list_skills now shows
  descriptions from skill frontmatter
- Kill the c_suite compile-all branch; everyone resolves bindings-only,
  capped at 200k chars with omissions named in the prompt
- Remove the Skills settings page and its import/sync/test IPC surface;
  drop ~/.agents discovery and the AGENTCOMPANY_NEW_SKILLS_ROOT seam in
  favor of AGENTCOMPANY_NEW_BUILTIN_SKILLS_DIR
- Tests: staging/import/forced-resolution/cap units, real-bundle CI
  check, e2e asserts the tab is gone and the catalogue self-imports
2026-07-08 21:18:50 -04:00

170 lines
8.9 KiB
TypeScript

import { randomUUID } from "node:crypto";
import { existsSync, mkdirSync, writeFileSync } from "node:fs";
import { tmpdir } from "node:os";
import { join } from "node:path";
import { describe, expect, test } from "vitest";
import { DatabaseConnection } from "../../src/main/database/connection";
import { OrganizationService } from "../../src/main/services/organization-service";
import { ChatService } from "../../src/main/services/chat-service";
import { SkillService, MAX_COMPILED_PROMPT_CHARS } from "../../src/main/services/skill-service";
import { importStagedSkills, parseFrontmatter, stageBuiltinSkillsLibrary } from "../../src/main/services/builtin-skills";
import { dispatchAgentTool, type AgentToolContext } from "../../src/main/services/agent-tools";
const HUMANIZER_MD = `---
name: humanizer
version: 2.1.1
description: |
Remove signs of AI-generated writing from text. Use when editing or reviewing
text to make it sound more natural and human-written.
allowed-tools:
- Read
---
# Humanizer
Write like a person: cut inflated symbolism and rule-of-three padding.
`;
const ARCHITECT_MD = `---
name: Backend Architect
description: Senior backend architect specializing in scalable system design.
color: blue
emoji: B
---
# Backend Architect Agent Personality
Design the systems that hold everything up.
`;
interface Harness {
connection: DatabaseConnection;
organization: OrganizationService;
skills: SkillService;
orgId: string;
sourceRoot: string;
libraryRoot: string;
}
function harness(): Harness {
const stamp = randomUUID();
const sourceRoot = join(tmpdir(), `ac-builtin-src-${stamp}`);
const libraryRoot = join(tmpdir(), `ac-builtin-lib-${stamp}`);
// Bundle fixture mirroring the real repo layout: a directory skill, a
// category of flat skills, and non-skill folders that must be ignored.
mkdirSync(join(sourceRoot, "humanizer"), { recursive: true });
writeFileSync(join(sourceRoot, "humanizer", "SKILL.md"), HUMANIZER_MD, "utf8");
writeFileSync(join(sourceRoot, "humanizer", "README.md"), "# About\n", "utf8");
mkdirSync(join(sourceRoot, "engineering"), { recursive: true });
writeFileSync(join(sourceRoot, "engineering", "engineering-backend-architect.md"), ARCHITECT_MD, "utf8");
writeFileSync(join(sourceRoot, "engineering", "README.md"), "docs, not a skill\n", "utf8");
mkdirSync(join(sourceRoot, "scripts"), { recursive: true });
writeFileSync(join(sourceRoot, "scripts", "install.sh"), "#!/bin/sh\n", "utf8");
const connection = new DatabaseConnection(join(tmpdir(), `ac-builtin-${stamp}.db`));
const organization = new OrganizationService(connection.db);
const skills = new SkillService(connection.db, join(tmpdir(), `ac-builtin-art-${stamp}`));
let orgId = "";
connection.transaction(() => {
orgId = organization.createOrganization({ organizationName: "OrbitOps", employeeName: "Yang", mission: "m", goal: "g" });
});
return { connection, organization, skills, orgId, sourceRoot, libraryRoot };
}
describe("built-in skills", () => {
test("stages both bundle shapes into the library and imports them once per org", () => {
const h = harness();
const staged = stageBuiltinSkillsLibrary(h.sourceRoot, h.libraryRoot);
expect(staged.map((unit) => unit.slug).sort()).toEqual(["engineering-backend-architect", "humanizer"]);
expect(existsSync(join(h.libraryRoot, "humanizer", "SKILL.md"))).toBe(true);
expect(existsSync(join(h.libraryRoot, "engineering-backend-architect", "SKILL.md"))).toBe(true);
const architect = staged.find((unit) => unit.slug === "engineering-backend-architect");
expect(architect?.manifest.name).toBe("Backend Architect");
expect(architect?.manifest.description).toContain("scalable system design");
const humanizer = staged.find((unit) => unit.slug === "humanizer");
expect(humanizer?.manifest.version).toBe("2.1.1");
expect(humanizer?.manifest.description).toContain("AI-generated writing");
const first = h.connection.transaction(() => importStagedSkills(h.skills, h.orgId, staged));
expect(first).toMatchObject({ imported: 2, skipped: 0, failures: [] });
const second = h.connection.transaction(() => importStagedSkills(h.skills, h.orgId, staged));
expect(second).toMatchObject({ imported: 0, skipped: 2 });
const slugs = (h.connection.db.prepare("select slug from skill_definitions where org_id = ? order by slug").all(h.orgId) as Array<{ slug: string }>)
.map((row) => row.slug);
expect(slugs).toEqual(["engineering-backend-architect", "humanizer"]);
h.connection.close();
});
test("every agent's run carries humanizer; hire-time skills join it and oversized grants are named, not overflowed", () => {
const h = harness();
// A third, oversized flat skill to prove the compiled-prompt cap.
writeFileSync(
join(h.sourceRoot, "engineering", "engineering-epic-novelist.md"),
`---\nname: Epic Novelist\ndescription: too big to pack\n---\n\n${"x".repeat(MAX_COMPILED_PROMPT_CHARS + 10)}\n`,
"utf8",
);
const staged = stageBuiltinSkillsLibrary(h.sourceRoot, h.libraryRoot);
h.connection.transaction(() => importStagedSkills(h.skills, h.orgId, staged));
let ceoId = "";
h.connection.transaction(() => {
const providerId = (h.connection.db.prepare("select id from providers where org_id = ? and provider_key = 'openai'").get(h.orgId) as { id: string }).id;
ceoId = h.organization.createAgent({
orgId: h.orgId, name: "CEO", role: "CEO", department: "Executive",
modelProviderId: providerId, modelId: "gpt-5.4", budgetCents: 25000, managerAgentId: null, roleLevel: "c_suite",
});
});
// The CEO hires with skills; the hire carries them plus forced humanizer.
const providerId = (h.connection.db.prepare("select model_provider_id from agents where id = ?").get(ceoId) as { model_provider_id: string }).model_provider_id;
const ceoCtx: AgentToolContext = {
orgId: h.orgId,
agent: { id: ceoId, modelProviderId: providerId, modelId: "gpt-5.4" },
taskProjectId: null,
taskGoalId: null,
workspace: null,
};
const deps = { db: h.connection.db, organization: h.organization, chat: new ChatService(h.connection.db) };
const hired = h.connection.transaction(() => dispatchAgentTool(deps, ceoCtx, {
id: "call_1",
name: "hire_agent",
arguments: { name: "Dana Lee", role: "Engineer", department: "Engineering", skills: ["engineering-backend-architect", "engineering-epic-novelist", "no-such-skill"] },
}));
expect(hired).toContain("They carry: engineering-backend-architect, engineering-epic-novelist");
expect(hired).toContain("Unknown skills skipped: no-such-skill");
const danaId = (h.connection.db.prepare("select id from agents where org_id = ? and name = 'Dana Lee'").get(h.orgId) as { id: string }).id;
const danaSnapshot = h.connection.transaction(() => h.skills.resolveRuntimeSnapshot({ orgId: h.orgId, agentId: danaId, runId: "run_dana" }));
const danaPrompt = (h.connection.db.prepare("select compiled_prompt from skill_runtime_snapshots where id = ?").get(danaSnapshot) as { compiled_prompt: string }).compiled_prompt;
expect(danaPrompt.indexOf("# Skill: humanizer")).toBe(0);
expect(danaPrompt).toContain("# Skill: Backend Architect");
expect(danaPrompt).toContain("Skills omitted for size");
expect(danaPrompt).toContain("Epic Novelist");
expect(danaPrompt.length).toBeLessThan(MAX_COMPILED_PROMPT_CHARS + 5_000);
// The CEO carries humanizer too - forced by resolution, not by binding.
expect(h.skills.hasRuntimeSkills(h.orgId, ceoId)).toBe(true);
const ceoSnapshot = h.connection.transaction(() => h.skills.resolveRuntimeSnapshot({ orgId: h.orgId, agentId: ceoId, runId: "run_ceo" }));
const ceoPrompt = (h.connection.db.prepare("select compiled_prompt from skill_runtime_snapshots where id = ?").get(ceoSnapshot) as { compiled_prompt: string }).compiled_prompt;
expect(ceoPrompt).toContain("# Skill: humanizer");
expect(ceoPrompt).not.toContain("# Skill: Backend Architect");
// list_skills shows descriptions and marks the house-style skill.
const listing = dispatchAgentTool(deps, ceoCtx, { id: "call_2", name: "list_skills", arguments: {} });
expect(listing).toContain("humanizer (humanizer)");
expect(listing).toContain("house style");
expect(listing).toContain("Backend Architect (engineering-backend-architect) — Senior backend architect");
h.connection.close();
});
test("frontmatter parsing reads inline values and block scalars", () => {
expect(parseFrontmatter(ARCHITECT_MD)).toMatchObject({
name: "Backend Architect",
description: "Senior backend architect specializing in scalable system design.",
});
expect(parseFrontmatter(HUMANIZER_MD)).toMatchObject({ name: "humanizer", version: "2.1.1" });
expect(parseFrontmatter(HUMANIZER_MD).description).toContain("sound more natural and human-written");
expect(parseFrontmatter("no frontmatter")).toEqual({});
});
});