io.github.Sahib-Sawhney-WH/looking-glass-mcp

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 "test_scenario_run" param "baseUrl" description: off-vendor URL in a description (localhost); 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 "test_auth_flow" has a secret-shaped input parameter "passwordSelector" — secrets must never be tool parameters (the model would supply/handle the credential). tool "test_auth_flow" has a secret-shaped input parameter "passwordValue" — secrets must never be tool parameters (the model would supply/handle the credential). tool "browser_login" has a secret-shaped input parameter "password" — secrets must never be tool parameters (the model would supply/handle the credential). tool "browser_login" has a secret-shaped input parameter "passwordSelector" — secrets must never be tool parameters (the model would supply/handle the credential).

• Permission scope

  pass15% of grade

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

  No shell-exec, unbounded/sensitive filesystem, or download-and-run primitives found in scanned source. PASS = no static red flag. Static scanning is blind to code fetched or executed at runtime.

• Version behavior

  fail10% of grade

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

  vs prior 3.1.0 (patch bump): new tool "browser_navigate" appeared; new tool "browser_go_back" appeared; new tool "browser_go_forward" appeared; new tool "browser_click" appeared; new tool "browser_type" appeared; new tool "browser_hover" appeared; new tool "browser_drag" appeared; new tool "browser_select_option" appeared; new tool "browser_press_key" touches a sensitive surface (email/exec/net/fs); new tool "browser_scroll" appeared; new tool "browser_screenshot" appeared; new tool "browser_snapshot" appeared; new tool "browser_console_messages" appeared; new tool "browser_network_requests" appeared; new tool "browser_evaluate" appeared; new tool "browser_tab_list" appeared; new tool "browser_tab_new" appeared; new tool "browser_tab_select" appeared; new tool "browser_close" appeared; new tool "test_scenario_run" touches a sensitive surface (email/exec/net/fs); new tool "test_scenario_status" appeared; new tool "test_assert" appeared; new tool "test_fill_form" appeared; new tool "test_auth_flow" appeared; new tool "test_watch_events" appeared; new tool "test_stop_watch" appeared; new tool "session_start" appeared; new tool "session_end" appeared; new tool "session_list" appeared; new tool "session_replay" appeared; new tool "visual_baseline" appeared; new tool "visual_compare" appeared; new tool "browser_resolve_element" appeared; new tool "browser_smart_click" appeared; new tool "browser_analyze_page" appeared; new tool "browser_wait_until_stable" appeared; new tool "browser_explore" appeared; new tool "browser_suggest_actions" appeared; new tool "browser_login" appeared; new tool "browser_fill_and_submit" appeared; new tool "browser_diagnose" appeared; new tool "browser_error_report" appeared; new tool "browser_get_cookies" appeared; new tool "browser_set_cookie" appeared; new tool "browser_get_localstorage" appeared; new tool "browser_set_localstorage" appeared; new tool "browser_clear_storage" appeared; new tool "browser_snapshot_state" appeared; new tool "browser_diff_state" appeared; new tool "browser_action_history" appeared; new tool "browser_clear_history" appeared; new tool "browser_health_check" appeared; new tool "browser_recover" appeared; new tool "browser_clean_slate" appeared; new tool "test_accessibility_audit" appeared; new tool "browser_performance_audit" appeared; new tool "test_generate_assertions" appeared; new tool "test_chaos" appeared; new tool "test_chaos_clear" appeared; new tool "session_export_playwright" appeared; new tool "browser_mock_route" appeared; new tool "browser_clear_mocks" appeared; new tool "browser_list_mocks" appeared; new tool "browser_extract" appeared; new tool "browser_go" appeared; new tool "browser_wait_for" appeared; new tool "browser_workflow" appeared; new tool "vault_list" appeared; new tool "vault_delete" touches a sensitive surface (email/exec/net/fs); new tool "vault_login" appeared; new tool "vault_inject" appeared.

• Transport config

  unscannable10% of grade

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

  local stdio server: no remote transport or auth surface, and no phone-home/telemetry signal found in source. Absence of transport findings is NOT a safety guarantee — phone-home hosts can be constructed at runtime and defeat static analysis.