Preflight Scout

RELEASE VERIFICATION FOR CODING AGENTS

Turn “done” into verifiable evidence.

Scout turns a diff into focused checks, runs them within explicit execution boundaries, and gives your agents, CI, and reviewers the evidence for release.

For AI agents: preflightscout.com/llms.txt

Install manually

Copy the setup prompt, then paste it into your coding agent.

Open sourceRuns in your repositoryExplicit execution boundaries

Before running the commands, confirm GitHub release v0.1.7 and the matching npm package both exist. If either is missing, use the source path on the install page.

  1. 1Install the CLI
    npm install --global @preflight-scout/cli@0.1.7 --registry=https://registry.npmjs.org/ preflight-scout install-browser
  2. 2Add the agent skill
    Codex
    codex plugin marketplace add fenutech/preflight-scout --ref plugin-stable codex plugin add preflight-scout@preflight-scout
    Claude Code
    claude plugin marketplace add fenutech/preflight-scout@plugin-stable claude plugin install preflight-scout@preflight-scout

    Restart the client and start a new task after installing either plugin. Folder installs are in the repository guide.

PS-01 LOCAL RUNREPO: fixture/checkoutBRANCH: feat/promo-total12:00:00Z

PRE-FLIGHT REPORT

RiskHIGH

DO NOT SHIP YET

Valid promo updates the total

PASSED
URL:
/
Viewport:
1280×720 Browser: Chromium
Evidence:
auto-valid-promo/final-observation.json

Expired promo preserves the total

FAILED
URL:
/
Viewport:
1280×720 Browser: Chromium
Expected:
$100.00 Observed: $90.00
Evidence:
auto-expired-promo/final-observation.json
Checks run: 2Passed: 1Failed: 1Blocked: 0

LOCAL EVIDENCE APPLY YOUR RELEASE POLICY

HOW IT WORKS

Plan for this change

Scout ties the analysis to your Git revisions and proposes checks for the affected flows. Your agent can review the plan.

Run within your boundaries

Agents execute within standing authorization. The built-in runner enforces reviewed actions, permissions, and the target origin.

Use the evidence

Screenshots, traces, failures, and final observations feed your agent, CI gate, or human review. Your release policy decides what follows.