A HAR exported "with sensitive data" holds the session cookie and Bearer token of the person who recorded it, and anyone who receives the file can open the site as that person without the password or the second factor. In 2023, session tokens taken from HAR files uploaded to a vendor's support system were used to hijack customers' admin sessions. This skill checks a HAR for ten kinds of sensitive data before you send or commit it, and tells you how to clean each one: export again sanitized, replace the values it names, and sign out so a copy that already went out stops working.

Your agent runs a scanner on your .har files (exports from Chrome, Edge, Firefox and Safari DevTools, Charles and Fiddler, and Playwright recordings used as test fixtures). It reads the HAR 1.2 structure entry by entry: request and response headers, cookies, URLs and query strings, post data (form, JSON, multipart), saved response bodies (base64 text included), redirects, WebSocket messages and page titles. It knows redacted values, deleted and analytics cookies, CSRF tokens and public browser keys, groups repeats per host, and reads files up to 100 MB. It never replays a request or sends the file anywhere. It returns BLOCK, REVIEW, or NOT FLAGGED, with the file, line, entry number, host, reason, and a safer approach. Secrets are never printed: findings name the header, cookie, parameter or field, and personal data is only counted.

What it checks

BLOCK (fix before shipping)

• H01: Authorization and Proxy-Authorization headers with a real credential, and JWTs in other headers

• H02: session cookies in Cookie and Set-Cookie headers and the HAR cookies lists

• H03: tokens, passwords, PHPSESSID and jsessionid session IDs, JWTs and signed-URL signatures (X-Amz-Signature, sig=) in URLs, redirects and Referer headers

• H04: passwords, client secrets, tokens, card numbers and JWTs in post data and sent WebSocket messages

• H05: access, refresh and ID tokens, passwords, private keys and JWTs in saved response bodies

REVIEW (check before release)

• H11: API keys in URLs, fields and headers, and X-Auth-Token, CSRF and other credential headers

• H12: one-time sign-in data (OAuth codes, PKCE verifiers, SAML responses, OTP codes)

• H13: email addresses and phone numbers in URLs, bodies, WebSocket messages and page titles

• H14: requests to private IP addresses and internal host names

• H15: captures over 20 MB or with more than 1,000 requests

What's included

• The scanner (Python 3.9+, standard library only)

• A rules guide with the reason and safer approach for each rule

• A risky and a safe example (a support capture and a test fixture), and the real report the scanner produces on the risky one

Limits

It reads the HAR as written and cannot tell whether a token is still valid. Bodies stored outside the HAR (Playwright's attach mode, .zip archives), other capture formats (.chls, .saz) and personal data other than emails and phone numbers are not checked. Files over 100 MB are listed as not checked. NOT FLAGGED is not a guarantee.

Contents

One ZIP file with 11 files: SKILL.md, the scanner in scripts/ (4 Python files), references/rules.md (every rule with its reason and safer approach), and examples/ (4 example files plus report.md, the real scanner output on the risky example).

How to use: This is an agent skill: a folder with a SKILL.md file and Python scripts that an AI coding agent such as Claude Code, Cursor, or Codex CLI can run. Unzip it into your agent's skills folder (for Claude Code: ~/.claude/skills/), then ask your agent to check your project. The scripts also run on their own with Python 3.9 or newer, with nothing to install. They read your files only: no network access, and nothing is sent anywhere.