io.github.duriantaco/skylos

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.

  skylos/llm/context.py: assigns a literal to API_KEY (sk-l…et (len 20)) — non-provider secret literal; prefer environment intake. test/test_mcp_rules.py: contains a hardcoded OpenAI API key (sk-a…no (len 30)) — in an example/sample file; verify it is not a real key. test/test_mcp_rules.py: contains a hardcoded Stripe live secret key (sk_l…no (len 35)) — in an example/sample file; verify it is not a real key. test/test_secrets.py: embeds a PRIVATE KEY block. (in an example/sample file — lower severity, but a real key here still leaks.) test/test_secrets.py: contains a hardcoded Google API key (AIza…AA (len 39)) — in an example/sample file; verify it is not a real key. test/test_typescript_expanded.py: assigns a literal to token (aB3d…45 (len 32)) — non-provider secret literal; prefer environment intake. Reads secrets from the environment (e.g. skylos/adapters/litellm_adapter.py) — the recommended intake shape.

• Permission scope

  warn15% of grade

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

  references credential paths: skylos/rules/quality/logic.py: references a credential path (id_rsa) has both shell/exec and network egress — capable of download-and-run, though no such path was observed (skylos/api.py: subprocess.* call; skylos/api.py: outbound HTTP/socket call). has both shell/exec and destructive filesystem writes (skylos/api.py: subprocess.* call; skylos/gatekeeper.py: open(…, 'w'/'a'/'x') write-mode file access). discloses shell/exec capability: 11 files (skylos/api.py, skylos/cicd/review.py, skylos/cicd/review.py, +8 more) discloses dynamic code execution: skylos/cicd/review.py: uses eval()/exec() discloses dynamic code execution: skylos/rules/danger/calls.py: uses eval()/exec() discloses dynamic code execution: skylos/visitors/languages/typescript/danger.py: uses eval()/exec() discloses broad/destructive filesystem access: skylos/gatekeeper.py: open(…, 'w'/'a'/'x') write-mode file access discloses broad/destructive filesystem access: skylos/rules/quality/logic.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) discloses broad/destructive filesystem access: skylos/sync.py: open(…, 'w'/'a'/'x') write-mode 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.duriantaco/[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

  warn10% of grade

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

  manifest/source references tracking behavior — disclosure