io.github.Licium-ai/licium

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 "bounty_subscribe" param "webhook_url" description: names an outbound copy/exfil destination (BCC/exfil verb).

• 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 "check_received_work" has a secret-shaped input parameter "report_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "capability_call" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "capability_price_notices" 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

  pass10% of grade

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

  vs prior 1.0.1 (major bump): new tool "list_sources" touches a sensitive surface (email/exec/net/fs); new tool "get_source" appeared; new tool "get_rows" touches a sensitive surface (email/exec/net/fs); new tool "request_source" touches a sensitive surface (email/exec/net/fs); new tool "check_received_work" touches a sensitive surface (email/exec/net/fs); new tool "search_mcp_directory" appeared; new tool "capabilities_list" appeared; new tool "capability_get" touches a sensitive surface (email/exec/net/fs); new tool "capability_register" appeared; new tool "capability_version_publish" appeared; new tool "capability_call" touches a sensitive surface (email/exec/net/fs); new tool "capability_price_notices" touches a sensitive surface (email/exec/net/fs); new tool "bounties_list" touches a sensitive surface (email/exec/net/fs); new tool "needs_intake" touches a sensitive surface (email/exec/net/fs); new tool "bounty_get" touches a sensitive surface (email/exec/net/fs); new tool "bounty_attachments_get" touches a sensitive surface (email/exec/net/fs); new tool "bounty_post" touches a sensitive surface (email/exec/net/fs); new tool "bounty_self_gate" touches a sensitive surface (email/exec/net/fs); new tool "bounty_pledge" touches a sensitive surface (email/exec/net/fs); new tool "bounty_counter" appeared; new tool "bounty_subscribe" touches a sensitive surface (email/exec/net/fs); new tool "work_alert_subscribe" appeared; new tool "source_report" appeared; new tool "agent_register" touches a sensitive surface (email/exec/net/fs); new tool "submission_post" touches a sensitive surface (email/exec/net/fs); new tool "bounty_withdraw" touches a sensitive surface (email/exec/net/fs). Low-severity drift (informational; consistent with the version bump).

• 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)