Pre-launch — Gluecron is in final validation. Public signups and git hosting for non-owner users open after launch review.
CodeIssuesDiscussionsWikiPull RequestsProjectsCommitsActionsReleasesContributorsPulse● GatesSecuritySettingsDeploymentsPipelineInsightsAgents✨ Explain✨ Ask AI✨ Workspace✨ Spec✨ Tests▓ Debt Map✨ NL Search🏛 Archaeology
claude/adoring-hopper-5x74bqclaude/affectionate-feynman-ykrf1hclaude/architecture-audit-design-wxprenclaude/build-status-update-3MXsfclaude/charming-meitner-mllb5rclaude/compare-gate-gluecron-s4mFQclaude/confident-faraday-tikcwbclaude/continue-work-XMTlIclaude/crontech-gluecron-deploy-7MIECclaude/crontech-platform-setup-SeKfwclaude/design-2026claude/ecstatic-ptolemy-jMdigclaude/enhance-github-integration-QNHdGclaude/fix-aa-loop-issue-PonMQclaude/fix-actions-and-processclaude/fix-desktop-errors-XqoW8claude/fix-red-workflowsclaude/fix-website-access-6FKJNclaude/gatetest-integration-hardeningclaude/github-audit-improvements-bDFr9claude/gluecron-launch-status-FoMRlclaude/hopeful-lamport-olfCTclaude/issue-to-pr-and-protectionsclaude/jolly-heisenberg-2sg1Qclaude/launch-preparation-QmTb6claude/new-session-xk1l7claude/plan-platform-architecture-kkN4yclaude/platform-analysis-roadmap-1nUGLclaude/platform-launch-assessment-8dWV8claude/polish-platform-release-AeDrUclaude/resume-previous-work-KzyLwclaude/review-crontech-handoff-qYEVqclaude/review-project-completeness-lHhS2claude/review-readme-docs-ulqPKclaude/serene-edison-rj87weclaude/setup-multi-repo-dev-BCwNQclaude/ship-fixes-and-tests-Jvz1cclaude/site-audit-competitive-pctlwgclaude/site-migration-vercel-XstpKclaude/standalone-product-repos-XHFTDcopilot/feat-smart-empty-states-keyboard-first-enhancementcopilot/feat-smart-morning-digest-review-context-restorecopilot/fix-and-process-workflowscopilot/update-ai-powered-code-reviewfeat/debt-mapfeat/push-policy-codeowners-hardeningfeat/smart-digest-contextfeat/stage-impactfeat/t1-secret-migrationfeat/u-polishfeat/w-self-hostfeat/w2-claude-configfix/agent-journey-orphan-sweepgatetest/auto-fix-1776586424172gatetest/auto-fix-1776586534814gatetest/auto-fix-1776590685143gatetest/auto-fix-1776590808199mainops/redeploy-retriggerstyle/dxt-cta-themeworktree-agent-a3377aad30d55da26worktree-agent-a7ef607b7ee1d6c74
gate-security-scan-availability.test.ts6.7 KB · 155 lines
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
/**
 * aiSecurityScanSafe() / gate.ts's runSecretAndSecurityScan — availability
 * vs verdict, mirroring gate-gatetest-availability.test.ts's pattern for
 * the exact same class of bug in a different gate.
 *
 * Before this fix, aiSecurityScan() returned `[]` on an AI-provider error
 * (timeout, rate limit, unparseable response) — indistinguishable from a
 * genuinely clean scan. gate.ts then reported "No security issues found",
 * silently converting a merge-blocking security gate into an automatic
 * pass on any Anthropic outage. Fixed the same way runGateTestScan's
 * outage bug was fixed (commit 6930df0): don't hard-block on a third-party
 * outage, but never claim "clean" when the scan didn't actually run.
 */

import { afterEach, beforeEach, describe, expect, it } from "bun:test";
import { runSecretAndSecurityScan } from "../lib/gate";
import { config } from "../lib/config";
import { __resetAnthropicClientForTests } from "../lib/ai-client";

const origFetch = globalThis.fetch;
const origKey = process.env.ANTHROPIC_API_KEY;

function mockFetch(responder: () => Response | Promise<Response>): void {
  // @ts-expect-error — override global fetch for the test
  globalThis.fetch = async (): Promise<Response> => responder();
}

function anthropicResponse(text: string): Response {
  return new Response(
    JSON.stringify({
      id: "msg_test",
      type: "message",
      role: "assistant",
      model: "claude-sonnet-4-6",
      content: [{ type: "text", text }],
      stop_reason: "end_turn",
      stop_sequence: null,
      usage: { input_tokens: 10, output_tokens: 10 },
    }),
    { status: 200, headers: { "content-type": "application/json" } }
  );
}

beforeEach(() => {
  process.env.ANTHROPIC_API_KEY = "sk-ant-test-key";
  // The Anthropic SDK client is a module-level singleton (ai-client.ts's
  // `_client`) that captures a fetch reference at construction time rather
  // than reading globalThis.fetch fresh per call — without resetting it,
  // only the FIRST test's globalThis.fetch mock would ever actually be
  // used, and every later test would silently reuse it (confirmed: this
  // leaked a 429-mock response into an unrelated later test before this
  // reset was added).
  __resetAnthropicClientForTests();
});

afterEach(() => {
  globalThis.fetch = origFetch;
  if (origKey === undefined) delete process.env.ANTHROPIC_API_KEY;
  else process.env.ANTHROPIC_API_KEY = origKey;
  __resetAnthropicClientForTests();
});

describe("runSecretAndSecurityScan — security scan availability vs verdict", () => {
  it("is skipped (not blocking) when the AI provider errors — was previously silently 'clean'", async () => {
    mockFetch(() => new Response("rate limited", { status: 429 }));
    const result = await runSecretAndSecurityScan("o", "r", "refs/heads/main", "a".repeat(40), {
      scanSecrets: false,
      scanSecurity: true,
      diffText: "diff --git a/x.ts b/x.ts\n+ eval(userInput)",
    });
    expect(result.securityResult.skipped).toBe(true);
    expect(result.securityResult.passed).toBe(true);
    expect(result.securityResult.details).not.toBe("No security issues found");
    expect(result.securityResult.details).toContain("unavailable");
  });

  it("is skipped (not blocking) when the request throws — network error / timeout", async () => {
    mockFetch(() => {
      throw new Error("ETIMEDOUT");
    });
    const result = await runSecretAndSecurityScan("o", "r", "refs/heads/main", "a".repeat(40), {
      scanSecrets: false,
      scanSecurity: true,
      diffText: "some diff",
    });
    expect(result.securityResult.skipped).toBe(true);
    expect(result.securityResult.passed).toBe(true);
    // The Anthropic SDK wraps a thrown-during-fetch error into its own
    // connection/timeout error type rather than preserving the original
    // message verbatim — assert on the outcome (skipped, not the clean
    // string), not the SDK's exact wording.
    expect(result.securityResult.details).not.toBe("No security issues found");
    expect(result.securityResult.details).toContain("unavailable");
  });

  it("is skipped (not blocking) when the AI response is unparseable", async () => {
    mockFetch(() => anthropicResponse("not json at all, just prose"));
    const result = await runSecretAndSecurityScan("o", "r", "refs/heads/main", "a".repeat(40), {
      scanSecrets: false,
      scanSecurity: true,
      diffText: "some diff",
    });
    expect(result.securityResult.skipped).toBe(true);
    expect(result.securityResult.passed).toBe(true);
    expect(result.securityResult.details).not.toBe("No security issues found");
  });

  // Regression guard: proves the two states are now actually distinguishable
  // — a genuine clean scan must still say "No security issues found" and
  // must NOT be marked skipped.
  it("is NOT skipped and reports a genuine clean verdict on success with zero findings", async () => {
    mockFetch(() => anthropicResponse('```json\n{"findings": []}\n```'));
    const result = await runSecretAndSecurityScan("o", "r", "refs/heads/main", "a".repeat(40), {
      scanSecrets: false,
      scanSecurity: true,
      diffText: "some diff",
    });
    expect(result.securityResult.skipped).toBe(false);
    expect(result.securityResult.passed).toBe(true);
    expect(result.securityResult.details).toBe("No security issues found");
  });

  it("blocks on a genuine finding — real verdict, not an outage", async () => {
    mockFetch(() =>
      anthropicResponse(
        '```json\n{"findings": [{"type": "Command Injection", "file": "x.ts", "line": 1, "severity": "critical", "description": "eval on user input"}]}\n```'
      )
    );
    const result = await runSecretAndSecurityScan("o", "r", "refs/heads/main", "a".repeat(40), {
      scanSecrets: false,
      scanSecurity: true,
      diffText: "some diff",
    });
    expect(result.securityResult.skipped).toBe(false);
    expect(result.securityResult.passed).toBe(false);
    expect(result.securityIssues).toHaveLength(1);
  });

  it("is skipped with 'no diff provided' when scanSecurity is requested but no diffText is given", async () => {
    const result = await runSecretAndSecurityScan("o", "r", "refs/heads/main", "a".repeat(40), {
      scanSecrets: false,
      scanSecurity: true,
    });
    expect(result.securityResult.skipped).toBe(true);
    expect(result.securityResult.passed).toBe(true);
    expect(result.securityResult.details).toBe("Skipped — no diff provided");
  });
});

// Keep config import referenced — sanity that config.anthropicApiKey reflects env for this test file.
describe("config.anthropicApiKey reflects ANTHROPIC_API_KEY for this suite", () => {
  it("resolves to the mocked value while set", () => {
    expect(config.anthropicApiKey).toBe("sk-ant-test-key");
  });
});