io.github.zapnikita95/agentpay

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 "begin_agent_link" description: URL on a cheap/abused TLD (agentpay.shop) — postmark-mcp used giftshop.club. tool "begin_agent_link" description: off-vendor URL in a description (agentpay.su); capability text should not name where data goes. tool "present_choices" description: off-vendor URL in a description (agentpay.su); capability text should not name where data goes. tool "create_purchase" description: imperative telling the model it must read/send/attach something.

• Supply chain

  pass25% of grade

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

  Python build script has no build-time network/exec; repository linkage verified

• 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

  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 0.2.1 (patch bump): tool "begin_agent_link" description changed and now contains embedded outbound URL; tool "verify_connection" description changed (wording only, no injection pattern); tool "verify_connection" gained input property(ies): sessionId; tool "get_balance" description changed (wording only, no injection pattern); tool "get_balance" gained input property(ies): sessionId; tool "get_limits" description changed (wording only, no injection pattern); tool "get_limits" gained input property(ies): sessionId; tool "get_recovery_guide" description changed (wording only, no injection pattern); tool "get_recovery_guide" gained input property(ies): sessionId; tool "get_spending_policy" description changed (wording only, no injection pattern); tool "get_spending_policy" gained input property(ies): sessionId; tool "list_allowed_stores" description changed (wording only, no injection pattern); tool "list_allowed_stores" gained input property(ies): sessionId; tool "get_user_preferences" description changed (wording only, no injection pattern); tool "get_user_preferences" gained input property(ies): sessionId; tool "update_preference" description changed (wording only, no injection pattern); tool "update_preference" gained input property(ies): sessionId; tool "get_delivery_address" description changed (wording only, no injection pattern); tool "get_delivery_address" gained input property(ies): sessionId; tool "save_delivery_address" description changed (wording only, no injection pattern); tool "save_delivery_address" gained input property(ies): sessionId; tool "peek_stores" description changed (wording only, no injection pattern); tool "peek_stores" gained input property(ies): sessionId; tool "search_products" description changed (wording only, no injection pattern); tool "search_products" gained input property(ies): query, search, need, text, sessionId; tool "get_product" description changed (wording only, no injection pattern); tool "get_product" gained input property(ies): sessionId; tool "present_choices" description changed and now contains embedded outbound URL; tool "present_choices" gained input property(ies): sessionId; tool "get_choice_status" description changed (wording only, no injection pattern); tool "get_choice_status" gained input property(ies): sessionId; tool "list_purchases" description changed (wording only, no injection pattern); tool "list_purchases" gained input property(ies): sessionId; tool "create_purchase" description changed (wording only, no injection pattern); tool "create_purchase" gained input property(ies): sessionId; tool "get_purchase_status" description changed (wording only, no injection pattern); tool "get_purchase_status" gained input property(ies): sessionId; tool "get_payment_status" description changed (wording only, no injection pattern); tool "get_payment_status" gained input property(ies): sessionId; tool "create_topup_intent" description changed (wording only, no injection pattern); tool "create_topup_intent" gained input property(ies): sessionId; tool "set_test_mode" description changed (wording only, no injection pattern); tool "set_test_mode" gained input property(ies): sessionId; tool "get_faq" description changed (wording only, no injection pattern); tool "get_faq" gained input property(ies): sessionId; tool "request_user_confirmation" description changed (wording only, no injection pattern); tool "request_user_confirmation" gained input property(ies): sessionId; new tool "get_agent_skills" touches a sensitive surface (email/exec/net/fs); new hardcoded outbound host/BCC destination(s): agentpay.shop, github.com.

• 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://agentpay.shop/ap/mcp" uses TLS (https); remote endpoint uses OAuth (strongest auth mode)