io.github.pretorin-ai/pretorin

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

  pass25% of grade

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

  repository linkage verified

• Credential hygiene

  warn15% of grade

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

  tests/test_agent_runner_coverage.py: assigns a literal to api_key (ptn-…ey (len 16)) — non-provider secret literal; prefer environment intake. tests/test_cli_deployment.py: assigns a literal to token (head…re (len 24)) — non-provider secret literal; prefer environment intake. tests/test_cli_deployment_coverage.py: embeds a PRIVATE KEY block. (in an example/sample file — lower severity, but a real key here still leaks.) tests/test_codex_agent.py: assigns a literal to api_key (ptn-…ey (len 16)) — non-provider secret literal; prefer environment intake. Reads secrets from the environment (e.g. src/pretorin/attestation.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 (src/pretorin/scanners/inspec.py: pipes a download into a shell). HIGH: executes content derived from a network response (src/pretorin/cli/deployment.py: passes a network response into an exec/eval call). has both shell/exec and destructive filesystem writes (src/pretorin/attestation.py: subprocess.* call; src/pretorin/plans.py: destructive filesystem call (shutil.rmtree/os.remove/etc.)). discloses shell/exec capability: 8 files (src/pretorin/attestation.py, src/pretorin/preflight_discovery.py, src/pretorin/agent/codex_runtime.py, +5 more) discloses broad/destructive filesystem access: 11 files (src/pretorin/plans.py, src/pretorin/preflight.py, src/pretorin/agent/codex_runtime.py, +8 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 0.28.10 (patch bump): new capability primitive(s) in source: child_process, fs-write, http-client, subprocess; new hardcoded outbound host/BCC destination(s): 127.0.0.1, alias.custom.example, api.example.com, api.openai.com, api.test, app.example.com, chatgpt.com, codex.legacy, configured.example, custom-openai.example, custom.example.com, default.example.com, developer.mozilla.org, different.com, docs.aws.amazon.com, env.example, env.test, evil.example, example.com, example.invalid, example.test, from-config.example, garage, github.com, harness.legacy, in-toto.io, learn.microsoft.com, legacy-codex.example, legacy-harness.example, legacy-stored.example, legacy.example, llm.example, mcp.example.com, model-primary.example, model.configured, model.env, model.env.example, model.setter, model.test, models.example, models.example.com, my-vllm.example.com, new.example, old.example, omnitruck.chef.io, other.com, plat.test, platform.custom.example, platform.example, platform.pretorin.com, platform.test, play.rust-lang.org, portal.test, pretorin-models.example, pretorin.com, proxy.example, pypi.org, raw.githubusercontent.com, schemas.openxmlformats.org, self-hosted.example.com, storage.example, stored-openai.example, test.pretorin.api, tracker.example, via-alias.example, www.w3.org.

• Transport config

  fail10% of grade

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

  source performs runtime self-update / download-and-execute (bypasses version_behavior); sends telemetry to third-party SDK (Datadog) — disclosure, verify opt-in; manifest/source references tracking behavior — disclosure