setup

name

setup

description

First-run setup for Browser-Data-Capture — provisions the user data folder under the standard $CLAUDE_USER_DATA pattern, optionally registers a path to the user's private maps repo for version-controlled storage, and verifies dependencies (mitmproxy, genson, pyyaml, optional openapi-spec-validator). Use when onboarding a new machine, when another skill reports a missing dependency, or when the user says "set up browser-data-capture" / "configure capture" / "where does this plugin store things".

setup

One-time / re-check setup for Browser-Data-Capture.

Steps

1. Resolve the data root.

   DATA_ROOT="${CLAUDE_USER_DATA:-${XDG_DATA_HOME:-$HOME/.local/share}/claude-plugins}/browser-data-capture"
   

   Create the standard layout if missing:

   $DATA_ROOT/
   ├── config.json     # small settings + pointer to private maps repo
   ├── data/           # per-capture run directories
   ├── cache/          # regenerable
   └── state/          # runtime state
   

2. Write a default config.json if absent:

   {
     "default_host_filter": null,
     "drop_static_assets": true,
     "redact_secrets": true,
     "private_maps_repo_path": null,
     "user_contact_for_disclosures": null
   }
   

   • redact_secrets should never be flipped to false from this skill — flag as manual-edit only.
   • private_maps_repo_path is a pointer to a user-owned local clone of a private GitHub repo (set up by init-private-repo). When set, create-domain-map writes maps there instead of into $DATA_ROOT/data/.
   • user_contact_for_disclosures (name + email) is used by disclose-finding to populate the "from" line in vulnerability disclosure drafts. Ask the user once during setup; never autofill.

3. Ask the user (interactive, one-time):

   • "Do you want me to set up a private GitHub repo for version-controlling your discovered endpoint maps?" — if yes, hand off to init-private-repo. If no, leave private_maps_repo_path as null and proceed.
   • "What name and email should I use if you ever need to send a disclosure email to a site owner?" — store in user_contact_for_disclosures. Skip if the user prefers to be prompted each time.

4. Check capture dependencies:

   • mitmproxy (mitmproxy --version) — primary capture path used by capture-via-proxy. If missing, give the install command (pip install --user mitmproxy or distro package) but don't auto-install.
   • HAR export is dependency-free (any browser does it) — note this as the zero-install path.
   • claude-in-chrome MCP server — optional. Only needed if the user wants observe-tab for live tab observation. Note as optional, not required.

5. Check inference dependencies:

   • genson — JSON schema inference (pip install --user genson).
   • pyyaml — YAML emission. Almost always present.
   • openapi-spec-validator — recommended, used by generate-openapi to validate output.

6. Report. Print a status table:

   data root:                <path>      [ok / created]
   config.json:              <path>      [ok / created]
   private maps repo:        <path|null>
   disclosure contact:       <name|null>
   mitmproxy:                <version|missing>
   claude-in-chrome MCP:     <present|absent>  (optional)
   genson:                   <version|missing>
   openapi-spec-validator:   <version|missing>  (recommended)
   

   For anything missing, give the exact install command and stop short of running it without confirmation.