io.github.ndl-systems/kevros

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 90 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

  warn25% of grade

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

  package name "kevros" is edit-distance 2 from popular package "keras" (typosquat candidate)

• Credential hygiene

  fail15% of grade

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

  tool "verify-token" has a secret-shaped input parameter "release_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "verify-token" has a secret-shaped input parameter "token_preimage" — secrets must never be tool parameters (the model would supply/handle the credential). Reads secrets from the environment (e.g. quickstart.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: kevros_governance/mcp_server.py: references a sensitive credential path (/mcp.json) — the tool-poisoning exfiltration target. discloses broad/destructive filesystem access: kevros_governance/mcp_server.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

  unscannable10% of grade

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

  no prior version to diff (first sight of io.github.ndl-systems/[email protected]); version-drift / rug-pull cannot be evaluated. Cold-start risk is covered by point-in-time checks (injection_surface, supply_chain, credential_hygiene), not here.

• 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)