@yadsh/dsh-web-fetch-authenticated

Configuration

Everything is editable from the DSH Web UI (Settings → Plugins → Authenticated Web Fetch): rules, credential write-only fields, network policy, redirects, limits, connection tester, and diagnostics. Declarative config stays available:

- id: web
  config:
    fetchProvider: authenticated

- id: web-fetch-authenticated
  name: '@yadsh/dsh-web-fetch-authenticated'
  config:
    rules:
      - id: corp-jira
        name: Corporate Jira
        enabled: true
        testUrl: https://jira.example.corp/status
        match:
          hosts: [jira.example.corp]
          allowPaths: [/browse/**, /rest/api/**]
        auth:
          type: bearer
          credential: JIRA_TOKEN        # credential REFERENCE, not the secret
        networkPolicy:
          allowPrivate: true
          allowedCidrs: [10.40.0.0/16]
        redirects:
          mode: same-origin
          maxRedirects: 3

- id: tool-web
  name: '@deepseek-ai/dsh-tool-web'
  config:
    fetch: true

Secrets are stored through the DSH credential store ($DSH_HOME credential backend) under POSIX-style reference names (JIRA_TOKEN); configuration keeps only the reference. Use the UI (or api.credentials.set) to store the value.

Security defaults

Policy	Default
Schemes	HTTPS only (per-rule http opt-in warns)
Private networks / loopback / link-local / CGNAT / IPv6 ULA	denied
Cloud metadata (169.254.169.254, fd00:ec2::254)	always denied
Redirects	same-origin only, max 3
Timeout / response size / body chars	30 s / 5 MiB / 100 k chars
Unmatched URLs	blocked (strict)

Narrow allowedCidrs are preferred over blanket allowPrivate: true. DNS answers are resolved, all candidates are classified, one denied answer denies the whole set (DNS-rebinding defense), and the socket is pinned to the approved addresses.