ai.sugra/api-mcp

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

  pass25% of grade

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

  Scanned 59 text field(s); no instruction-injection patterns found. PASS = no static red flag. A server can serve different tool descriptions at runtime (the rug-pull class), which static scanning cannot see.

• Supply chain

  fail25% of grade

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

  Python build script executes at build time: network call in build script

• Credential hygiene

  warn15% of grade

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

  tests/test_auth.py: assigns a literal to api_key (sugr…ey (len 16)) — non-provider secret literal; prefer environment intake. Reads secrets from the environment (e.g. evals/agent_eval.py) — the recommended intake shape.

• Permission scope

  fail15% of grade

  Declared tools vs. breadth (filesystem, network, exec); flags shell-exec and unbounded filesystem access.

  HIGH: shell/exec capability AND network egress together — download-and-execute risk (tests/test_agent_tools.py: subprocess.* call; evals/smoke_live.py: outbound HTTP/socket call). HIGH: shell/exec capability AND destructive filesystem writes together (tests/test_agent_tools.py: subprocess.* call; scripts/build_endpoint_catalog.py: path traversal ('../') combined with file access). HIGH: shell/exec capability is incongruent with a read-only declared purpose (tools look like lookup/read-only but source imports command execution). discloses shell/exec capability: tests/test_agent_tools.py: subprocess.* call discloses shell/exec capability: tests/test_cli.py: subprocess.* call discloses shell/exec capability: tests/test_metadata_sync.py: subprocess.* call discloses broad/destructive filesystem access: scripts/build_endpoint_catalog.py: path traversal ('../') combined with file access Capability DISCLOSURE, not a verdict: static analysis sees the primitive is present and reachable, not whether its use is attacker-controlled.

• 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 0.9.0 (patch bump): new tool "sugra_entity_screen" appeared; new tool "sugra_entity_lookup" appeared; new tool "search_endpoints" appeared; new tool "describe_endpoint" touches a sensitive surface (email/exec/net/fs); new tool "call_endpoint" appeared; new tool "list_toolsets" appeared; new tool "fetch_data" touches a sensitive surface (email/exec/net/fs); new tool "list_sources" appeared; new tool "resolve_entity" appeared; new tool "get_snapshot" appeared; new tool "get_timeseries" appeared; new capability primitive(s) in source: http-client, subprocess; new hardcoded outbound host/BCC destination(s): 127.0.0.1, api.test, app.cursor.sh, app.sugra.ai, attacker.example, chat.openai.com, chatgpt.com, claude.ai, claude.com, cursor.sh, custom.example, evil.example, example.com, other.example, platform.openai.com, x.

• 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); manifest/source references telemetry behavior — disclosure