io.agent4/agent4-tenant

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

  fail25% of grade

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

  tool "improve_prompt" description: text addressed to the model rather than describing a capability. tool "set_custom_domain" description: off-vendor URL in a description (chat.client.com); capability text should not name where data goes.

• Supply chain

  warn25% of grade

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

  no resolvable repository linked to the package (provenance unverifiable)

• Credential hygiene

  pass15% of grade

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

  No hardcoded secrets, secret-shaped tool parameters, or leaked keys found. PASS = no static red flag; static analysis cannot prove the code honors env-based secret handling at runtime.

• 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 1.3.0 (minor bump): tool "list_agents" description changed (wording only, no injection pattern); tool "get_agent" description changed (wording only, no injection pattern); tool "create_agent" description changed (wording only, no injection pattern); tool "create_agent" gained input property(ies): ask_forms; tool "update_agent" description changed (wording only, no injection pattern); tool "update_agent" gained input property(ies): add_tools, remove_tools, add_skills, remove_skills, add_knowledge_bases, remove_knowledge_bases, auto_suggest, suggest_prompt; tool "list_tools" description changed (wording only, no injection pattern); tool "list_mcp_servers" description changed (wording only, no injection pattern); tool "create_share" description changed (wording only, no injection pattern); tool "list_shares" description changed (wording only, no injection pattern); tool "configure_share" description changed (wording only, no injection pattern); tool "configure_share" gained input property(ies): label, highlight_color; tool "set_pwa_branding" description changed (wording only, no injection pattern); tool "set_pwa_branding" gained input property(ies): agent; tool "set_custom_domain" description changed (wording only, no injection pattern); tool "list_skills" description changed (wording only, no injection pattern); tool "get_skill" description changed (wording only, no injection pattern); tool "create_skill" description changed (wording only, no injection pattern); tool "update_skill" description changed (wording only, no injection pattern); tool "update_skill" gained input property(ies): add_tools, remove_tools; tool "list_knowledge_bases" description changed (wording only, no injection pattern); tool "get_knowledge_base" description changed (wording only, no injection pattern); tool "create_knowledge_base" description changed (wording only, no injection pattern); tool "create_knowledge_base" gained input property(ies): top_k; tool "update_knowledge_base" description changed (wording only, no injection pattern); tool "update_knowledge_base" gained input property(ies): top_k; tool "add_knowledge_text" description changed (wording only, no injection pattern); tool "add_knowledge_file" description changed (wording only, no injection pattern); tool "search_knowledge_base" description changed (wording only, no injection pattern); tool "improve_prompt" description changed (wording only, no injection pattern); tool "diagnose_prompt" description changed (wording only, no injection pattern); tool "usage_stats" description changed (wording only, no injection pattern); tool "tenant_info" description changed (wording only, no injection pattern); tool "list_end_users" description changed (wording only, no injection pattern); tool "get_end_user" description changed (wording only, no injection pattern); tool "list_user_sessions" description changed (wording only, no injection pattern); tool "list_page_contexts" description changed (wording only, no injection pattern); tool "upsert_page_context" description changed (wording only, no injection pattern); tool "resolve_page_context" description changed (wording only, no injection pattern); tool "page_context_stats" description changed (wording only, no injection pattern); tool "list_storylines" description changed (wording only, no injection pattern); tool "get_storyline" description changed (wording only, no injection pattern); tool "create_storyline" description changed (wording only, no injection pattern); tool "create_storyline" gained input property(ies): user_visibility, concurrency, entry, display_name, allow_exit, show_profile; tool "update_storyline" description changed (wording only, no injection pattern); tool "update_storyline" gained input property(ies): user_visibility, entry, display_name, allow_exit, show_profile, concurrency; tool "validate_storyline" description changed (wording only, no injection pattern); tool "publish_storyline" description changed (wording only, no injection pattern); tool "unpublish_storyline" description changed (wording only, no injection pattern); tool "clone_storyline" description changed (wording only, no injection pattern); tool "export_storyline" description changed (wording only, no injection pattern); tool "import_storyline" description changed (wording only, no injection pattern); new tool "test_skill_trigger" touches a sensitive surface (email/exec/net/fs); new tool "build_knowledge_index" touches a sensitive surface (email/exec/net/fs); new tool "get_knowledge_index" appeared; new tool "patch_knowledge_index" appeared; new tool "test_planner_trigger" touches a sensitive surface (email/exec/net/fs); new tool "preview_plan" 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)