com.watchtower-api/sanctions-screening

What we checked

Six static checks, weighted by risk. Every result reflects only what could be observed in the published package and repository — never intent.

• Injection surface

  warn25% of grade

  Tool descriptions/manifest scanned for instruction-injection patterns (imperatives at the model, hidden text, 'ignore previous', data-exfil URLs).

  tool "screen_name" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "screen_names_batch" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "screen_crypto_address" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "create_watch" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "validate_identifier" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "check_jurisdiction" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "scan_text" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "resolve_lei" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "search_lei_by_name" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes. tool "kyb_check" param "api_key" description: off-vendor URL in a description (watchtower-api.com); capability text should not name where data goes.

• Supply chain

  pass25% of grade

  Package provenance: namespace verification, repo linkage, maintainer count, account age, postinstall scripts, typosquat distance.

  repository linkage verified

• Credential hygiene

  fail15% of grade

  How the server takes secrets (env vs plaintext config vs hardcoded); secrets appearing in tool schemas.

  tool "screen_name" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "screen_names_batch" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "screen_crypto_address" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_watch" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "list_watches" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "validate_identifier" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "check_jurisdiction" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "scan_text" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "resolve_lei" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "search_lei_by_name" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "kyb_check" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential).

• Permission scope

  unscannable15% of grade

  Declared tools vs. breadth (filesystem, network, exec); flags shell-exec and unbounded filesystem access.

  No source files, package.json, or pyproject were fetched; cannot infer capability scope.

• Version behavior

  warn10% of grade

  Diff of tool definitions between versions; new permissions or changed descriptions in a patch release (the postmark-mcp class).

  vs prior 1.0.0 (minor bump): new tool "create_api_key" touches a sensitive surface (email/exec/net/fs); new tool "screen_name" touches a sensitive surface (email/exec/net/fs); new tool "screen_names_batch" touches a sensitive surface (email/exec/net/fs); new tool "screen_crypto_address" appeared; new tool "create_watch" touches a sensitive surface (email/exec/net/fs); new tool "list_watches" appeared; new tool "validate_identifier" touches a sensitive surface (email/exec/net/fs); new tool "check_jurisdiction" appeared; new tool "scan_text" touches a sensitive surface (email/exec/net/fs); new tool "resolve_lei" appeared; new tool "search_lei_by_name" appeared; new tool "kyb_check" appeared; new tool "get_screening_info" appeared.

• Transport config

  fail10% of grade

  Remote servers: TLS and auth mode (none/token/OAuth). Local servers: whether the manifest indicates it phones home.

  remote MCP endpoint declares NO authentication (unauthenticated remote — nginx-ui/CVE-2026-33032 class)