- Home
- Skills
- Legal & Compliance
- last-look
Last Look
A first-pass check, not legal advice, for common legal, app store and good-practice mistakes in web, Android, iOS and cross-platform projects or live sites. India in depth; EU, UK, US, Apple and Google Play in part.
Free
Works with the AI Agents you already use
last-look
Example session with this skill installed
Use last-look on this Expo app. The business is based in India and we want users only in India. We'll publish on Google Play and the Apple App Store. It's a hiking notes app for adults, not for children. We don't sell user data; I'm not sure what else counts as sharing. There's no website.
- Read your context and instructions
- Compiled the last-look
- Generated the UI component
I've written the report to last-look-report.md at the project root. It lists 6 findings at Now, 5 at Soon and none at Later. Four are confirmed in the code and seven are marked "check on your own".
Top Now items
- No privacy policy link: nothing in the app points to one, and both stores expect it (
App.tsx). - Location logged:
App.tsx:17writes the exact coordinates to the console.
Android target level: nothing sets it, and Expo 51 likely defaults below the API 36 that Google Play has required since 31 August 2026. Check your latest build.
Very little functionality: the only screen shows your coordinates. There is no note, photo or map code, so reviewers may reject it.
- Store privacy forms: these live outside the repo, so declare precise location and anything you add later.
Breach plan: CERT-In (India's cyber-incident agency) expects a plan for reporting cyber incidents within 6 hours. None was found.
Your "not sure what counts as sharing" answer: sharing means sending personal data off the device to anyone, not only selling it. That includes analytics, crash reporting, push services, backups and ad SDKs. The code shows none of these today, but the store forms and your policy need to cover anything you add.
What the report does not cover
- Search for new rules: it did not run, because no web tool was available.
- EU, UK and US rules: their files weren't opened, since you only target India.
- Live-site checks: skipped, because there is no URL.
- Security: it is not a full audit, so run a secret scanner and a dependency audit as well.
The report may contain sensitive project details, so check it before you share or commit it. I can fix specific findings if you ask. The simplest ones are the log line, the launch permission requests and the unused camera and notification permissions.
App.tsx
TSX · React component
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
A first-pass check, not legal advice, for common legal, app store and good-practice mistakes in web, Android, iOS and cross-platform projects or live sites. It can never confirm compliance: a clean report means only that these checks found nothing. For decisions that matter, ask a qualified lawyer.
What it does
last-look is plain Markdown instructions that your AI coding agent follows. There is no code to run. Point your agent at a project folder, a live URL or both, and it:
- Looks at the project and asks only what it could not tell: where the business is based, which countries you want users in, which app stores, whether children use it, and whether you sell or share data. "I don't know" is a valid answer and is reported as an open risk.
- Works out which rule sets probably apply, and loads only the checks the project needs.
- Runs the checks, then a critic pass that removes findings with no location or evidence.
- Searches the web for newer rules if the agent has a web tool. These go in their own "unverified" section.
- Writes
last-look-report.mdto the project folder.
Each finding gives:
- the location, as
file:linewhere possible - how sure it is: confirmed in code, confirmed on the live site, or "Maybe: check on your own"
- why it matters, severity and priority (Now, Soon or Later)
- a fix where one is possible
- the rule behind it, with its strength (law, court or regulator, store rule, or good practice), when the rule started, its status, when it was last checked against the source, and a source link
Good-practice findings sit in their own section, so they never look like law. Say "rules only" to leave them out, or "no search" to skip the web search.
What it covers
93 checks in 19 areas and 248 rule records.
| Rules | Records | How deep |
|---|---|---|
| India | 97 | DPDP Act and Rules, IT Rules 2021, CERT-In Directions 2022, E-Commerce Rules 2020 and Dark Patterns Guidelines 2023 fully walked; RBI recurring-payment rules targeted |
| Google Play | 48 | Developer Program Policy fully walked |
| Apple | 36 | App Review Guidelines fully walked |
| Good practice | 31 | Apple and Android privacy docs, OWASP, WCAG 2.2, open-source licences |
| EU and UK | 24 | Targeted: GDPR, ePrivacy, UK GDPR and PECR, European Accessibility Act |
| US | 12 | Targeted: COPPA, CAN-SPAM, CCPA, CIPA, California auto-renewal law, DMCA, CalOPPA s.22575 |
A full walk means every section of the source has a recorded decision. A targeted walk covers only the sections the checks rely on. Countries with no rules file get the common checks and web search results, and the report says no list exists yet.
The demo is a real run on expo-bad, a made-up test project with planted mistakes from the GitHub repo.
Source, issues and test results: https://github.com/KrunalSM/last-look
- It reads code and, with a browser tool, one live site. It can't see app store consoles, ad or analytics settings, contracts, server setup or what staff do. Those come out as "Maybe: check on your own".
- It is not a security scanner. The report points to dedicated security tools.
- Health, finance, gambling and similar fields are flagged as needing extra rules, not checked in depth.
- Live sites: public pages only. No logging in, purchases or form submissions.
- Rules change. Each record shows when it was last checked against its source, and the report lists records past their check date.
- The agent can make mistakes. Re-read each cited line before acting on it.
- It is not legal advice and never confirms compliance.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed today
- Free with an Agensi account
Permissions
File Scopes
Reads the project folder. Writes only last-look-report.md at the project root unless you ask for fixes. Live sites: public pages only, with no logins, purchases or form submissions. Web search queries contain legal topics only, never project code or details. last-look itself collects and sends nothing.