io.github.clayseal/clayseal

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: subprocess in build script, 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.

  benchmarks/flow_probes.py: assigns a literal to secret (AKIA…QZ (len 22)) — non-provider secret literal; prefer environment intake. benchmarks/bpl/scenarios/edgecases.py: assigns a literal to SECRET (APIK…3a (len 16)) — non-provider secret literal; prefer environment intake. benchmarks/tests/test_flow_invariants.py: assigns a literal to SECRET (AKIA…QZ (len 22)) — non-provider secret literal; prefer environment intake. python/tests/test_bounded_work_fails_closed.py: assigns a literal to SECRET (AKIA…QZ (len 22)) — non-provider secret literal; prefer environment intake. python/tests/test_confidentiality.py: assigns a literal to SECRET (AKIA…QZ (len 22)) — non-provider secret literal; prefer environment intake. python/tests/test_decode_layer.py: contains a hardcoded AWS access key id (AKIA…4X (len 20)) — in an example/sample file; verify it is not a real key. Reads secrets from the environment (e.g. benchmarks/aml_sequence_validation.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: 7 files (benchmarks/adversarial/adaptive.py, benchmarks/adversarial/attacks.py, benchmarks/datasets/ipi_coding.py, +4 more) has both shell/exec and network egress — capable of download-and-run, though no such path was observed (benchmarks/gateway_cost.py: subprocess.* call; benchmarks/_http.py: outbound HTTP/socket call). has both shell/exec and destructive filesystem writes (benchmarks/gateway_cost.py: subprocess.* call; benchmarks/adaptive_decay.py: open(…, 'w'/'a'/'x') write-mode file access). discloses shell/exec capability: 8 files (benchmarks/gateway_cost.py, benchmarks/verify_results.py, benchmarks/live/frontier.py, +5 more) discloses dynamic code execution: clayseal/capabilities/monitor/entailment.py: uses eval()/exec() discloses dynamic code execution: clayseal/capabilities/monitor/scoring/transformer.py: uses eval()/exec() discloses broad/destructive filesystem access: 17 files (benchmarks/adaptive_decay.py, benchmarks/bpl_ordering_control.py, benchmarks/invariance_rungs.py, +14 more) 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.clayseal/[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.

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