vigilis

The trust layer for autonomous testing. An AI agent that writes, gates, triages, and self-heals your Playwright, Cypress, and Selenium tests — and seals every run in a signed, independently verifiable receipt. Self-healing is the wedge; verifiable proof is the point.

🌐 vigilis.dev · 💻 GitHub

Install

npm i -D vigilis        # in your Playwright, Cypress, or Selenium project
npx playwright install chromium   # one-time, if you haven't (needed for browser automation)
export ANTHROPIC_API_KEY=sk-...

Needs a real Anthropic API key. Brings its own Playwright; uses your project's specs.

Use

vigilis init                              # scaffold vigilis.config.json (baseUrl, testDir, model, framework)
vigilis generate https://your-app.com --run                        # explore → write + run a real spec (auto-detects framework)
vigilis generate https://your-app.com --framework cypress --run    # force Cypress output
vigilis triage  https://your-app.com --spec tests/login.spec.ts   # real-bug vs DOM drift vs flake
vigilis heal    https://your-app.com --spec tests/login.spec.ts   # heal drift → verify green → PR
vigilis attest-run report.json --commit $GITHUB_SHA --exit-code $?  # seal any test run into a receipt (no key needed)
vigilis verify  .vigilis/attestation/qa-run-<sha>.json              # re-walk the chain: intact or tampered

• Generate real specs from a URL — Playwright (most battle-tested), Cypress, or Selenium. Framework is auto-detected from your project; pass --framework playwright|cypress|selenium to force one.
• Gate them in CI — failing tests block the deploy.
• Triage failures: real bug vs DOM drift vs flake.
• Heal cosmetic drift (rewrite locator → re-verify green → open a PR) — and refuse to heal a real bug, so it never papers a regression into a false green.
• Attest any test run — vigilis attest-run hash-chains a Playwright JSON report's digest, the commit SHA, and the runner's exit code into a tamper-evident receipt. Zero secrets, zero API calls — it works in any CI job with no key configured.
• Verify a receipt offline — vigilis verify <bundle> re-walks the hash chain and tells you intact or broken at record #N. Anyone with the file can run it; no account, no network.

Receipts

Every attested run produces a hash-chained bundle: each record embeds the hash of the previous one, so editing or removing any step breaks every hash after it. Receipts prove the recorded steps are unaltered — they are verifiable and auditable, not a correctness guarantee.

• Default (zero-secret): local bundles under .vigilis/attestation/, checked with vigilis verify. Works everywhere, including air-gapped CI.
• Signed (optional): when the Treeship CLI is installed, every heal is additionally sealed into a signed receipt by an independent notary — verify with treeship verify last, or share the hosted URL. No hard dependency; --no-receipt to opt out.

vigilis.config.json supplies per-project defaults (baseUrl, testDir, model, framework); explicit flags always override it.

Drive it from Claude (MCP)

The same tools ship as an MCP server — generate/triage/heal straight from Claude Desktop/Code. See the docs.

MIT © Piyush Pathak · provenance powered by Treeship (Zerker Lab)