dev.voicemode/voicemode

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

  unscannable25% of grade

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

  No tool descriptions, server instructions, prompts, or resources were fetched; nothing to scan for injection.

• 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.

  voice_mode/serve_middleware.py: assigns a literal to secret (my-s…id (len 14)) — non-provider secret literal; prefer environment intake. voice_mode/serve_middleware.py: assigns a literal to token (my-s…en (len 15)) — non-provider secret literal; prefer environment intake. Reads secrets from the environment (e.g. voice_mode/_env_deprecation.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: downloads and pipes straight to a shell (voice_mode/tools/kokoro/install.py: pipes a download into a shell). HIGH: executes content derived from a network response (voice_mode/utils/version_helpers.py: passes a network response into an exec/eval call). has both shell/exec and destructive filesystem writes (voice_mode/cli.py: subprocess.* call; voice_mode/config.py: open(…, 'w'/'a'/'x') write-mode file access). discloses shell/exec capability: 34 files (voice_mode/cli.py, voice_mode/conch_notify.py, voice_mode/conch_ops.py, +31 more) discloses broad/destructive filesystem access: 31 files (voice_mode/config.py, voice_mode/config.py, voice_mode/conversation_logger.py, +28 more) 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 8.11.0 (minor bump): new capability primitive(s) in source: fs-write, http-client, net, subprocess; new hardcoded outbound host/BCC destination(s): 127.0.0.1, api.cartesia.ai, api.github.com, api.openai.com, astral.sh, brew.sh, cartesia.ai, docs.anthropic.com, download.pytorch.org, example.com, getvoicemode.com, github.com, host, huggingface.co, mpv.io, ms2, musicforprogramming.net, platform.openai.com, play.cartesia.ai, raw.githubusercontent.com, stream.example.com, voicemode.dev, www.ffmpeg.org.

• Transport config

  warn10% of grade

  Remote servers: TLS and auth mode (none/token/OAuth). Local servers: whether the manifest indicates it phones home.

  sends telemetry to third-party SDK (amplitude) — disclosure, verify opt-in; manifest/source references heartbeat behavior — disclosure