ai.physea/liminality

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).

  server instructions: off-vendor URL in a description (physea.ai); 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

  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

  fail10% of grade

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

  vs prior 1.0.4 (patch bump): new tool "solve" appeared; new tool "get_capabilities" appeared; new tool "get_settings" touches a sensitive surface (email/exec/net/fs); new tool "set_settings" appeared; new tool "get_result_detail" touches a sensitive surface (email/exec/net/fs); new tool "export_result" appeared; new tool "compress_context" appeared; new tool "delegation" appeared; new tool "handoff" appeared; new tool "ask_form" appeared; new tool "apply_form" appeared; new tool "report_outcome" appeared; new tool "report_feedback" appeared; new tool "apply_grade" appeared; new tool "optimize" appeared; new tool "algorithms" touches a sensitive surface (email/exec/net/fs); new tool "task" touches a sensitive surface (email/exec/net/fs); new tool "recover" touches a sensitive surface (email/exec/net/fs); new tool "runs" touches a sensitive surface (email/exec/net/fs); new tool "research" touches a sensitive surface (email/exec/net/fs); new tool "composio_connect" appeared; new tool "register_asset" appeared; new tool "set_priorities" appeared; new tool "taxis" appeared; new tool "catalog" touches a sensitive surface (email/exec/net/fs); new tool "setup" appeared; new tool "manage" 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://liminality.physea.ai/mcp" uses TLS (https); remote endpoint "https://mcp.physea.ai/mcp" uses TLS (https); remote endpoint uses OAuth (strongest auth mode)