ai.finestructure/fine-structure

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 "read_app_file" description: off-vendor URL in a description (finestructure.ai); capability text should not name where data goes. tool "read_app_files" description: off-vendor URL in a description (finestructure.ai); capability text should not name where data goes. tool "write_app_file" description: off-vendor URL in a description (finestructure.ai); capability text should not name where data goes. tool "patch_app_file" description: off-vendor URL in a description (finestructure.ai); capability text should not name where data goes. tool "add_file_change" description: off-vendor URL in a description (finestructure.ai); capability text should not name where data goes. tool "restore_file_from_version" description: off-vendor URL in a description (finestructure.ai); 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

  fail15% of grade

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

  tool "invite_app_member" has a secret-shaped input parameter "password" — 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

  fail10% 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 (patch bump): tool "read_app_file" description changed and now contains embedded outbound URL; tool "read_app_files" description changed and now contains embedded outbound URL; tool "write_app_file" description changed and now contains embedded outbound URL; tool "patch_app_file" description changed and now contains embedded outbound URL; tool "seed_entity" description changed (wording only, no injection pattern); tool "create_entity_records" description changed (wording only, no injection pattern); tool "update_entity_record" description changed (wording only, no injection pattern); tool "update_entity_records" description changed (wording only, no injection pattern); tool "delete_entity_records" description changed (wording only, no injection pattern); tool "add_file_change" description changed and now contains embedded outbound URL; tool "invite_app_member" description changed (wording only, no injection pattern); tool "restore_file_from_version" description changed and now contains embedded outbound URL; new tool "agency_create_client" appeared; new tool "agency_list_clients" appeared; new tool "agency_create_client_app" touches a sensitive surface (email/exec/net/fs); new tool "agency_create_claim_link" touches a sensitive surface (email/exec/net/fs); new tool "agency_get_claim_status" appeared; new tool "set_app_subdomain" appeared; new tool "check_subdomain_available" appeared; new tool "get_app_analytics" appeared; new tool "create_agent" touches a sensitive surface (email/exec/net/fs); new tool "list_agents" appeared; new tool "schedule_agent_task" touches a sensitive surface (email/exec/net/fs); new tool "list_agent_tasks" touches a sensitive surface (email/exec/net/fs); new tool "get_agent_whatsapp" touches a sensitive surface (email/exec/net/fs); new tool "start_owner_phone_verification" touches a sensitive surface (email/exec/net/fs); new tool "confirm_owner_phone_verification" appeared; new tool "attach_agent_whatsapp" appeared; new tool "detach_agent_whatsapp" appeared.

• Transport config

  pass10% of grade

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

  remote endpoint "https://finestructure.ai/api/mcp" uses TLS (https); remote endpoint uses OAuth (strongest auth mode)