ai.geodesiclabs/governance-platform

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

  warn25% of grade

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

  server instructions: off-vendor URL in a description (app.geodesiclabs.ai); capability text should not name where data goes.

• Supply chain

  warn25% of grade

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

  no resolvable repository linked to the package (provenance unverifiable)

• Credential hygiene

  fail15% of grade

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

  tool "validate" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "validate_repair" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_blueprint" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "list_blueprints" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "repair" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "check_blueprint_health" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "compare_semantic_equivalence" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "govern_inference" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_inference_trace" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "recent_inference_decisions" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "verify_certificate" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "profile_blueprint_robustness" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "forecast" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "discover_patterns" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "repair_path" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "counterfactual" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "analyze_anomaly" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_chain" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "submit_chain_stage" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "handoff_audit" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "approve_rule" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "reject_rule" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "structural_types" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "decompose_failure" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "geometric_confidence" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "check_realization" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "check_drift" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "authorize_execution" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "load_rule_pack" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_execution_trace" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "verify_replay" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "account_status" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "delete_blueprint" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "update_blueprint" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "list_api_keys" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "rotate_api_key" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential). tool "delete_api_key" has a secret-shaped input parameter "api_key" — secrets must never be tool parameters (the model would supply/handle the credential).

• Permission scope

  unscannable15% of grade

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

  No source files, package.json, or pyproject were fetched; cannot infer capability scope.

• Version behavior

  pass10% of grade

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

  vs prior 1.0.0 (minor bump): tool "validate" description changed (wording only, no injection pattern); tool "validate_repair" description changed (wording only, no injection pattern); tool "create_blueprint" description changed (wording only, no injection pattern); tool "list_blueprints" description changed (wording only, no injection pattern); tool "repair" description changed (wording only, no injection pattern); tool "check_blueprint_health" description changed (wording only, no injection pattern); tool "compare_semantic_equivalence" description changed (wording only, no injection pattern); tool "govern_inference" description changed (wording only, no injection pattern); tool "get_inference_trace" description changed (wording only, no injection pattern); tool "recent_inference_decisions" description changed (wording only, no injection pattern); tool "verify_certificate" description changed (wording only, no injection pattern); tool "profile_blueprint_robustness" description changed (wording only, no injection pattern); tool "forecast" description changed (wording only, no injection pattern); tool "discover_patterns" description changed (wording only, no injection pattern); tool "repair_path" description changed (wording only, no injection pattern); tool "counterfactual" description changed (wording only, no injection pattern); tool "analyze_anomaly" description changed (wording only, no injection pattern); tool "create_chain" description changed (wording only, no injection pattern); tool "submit_chain_stage" description changed (wording only, no injection pattern); tool "handoff_audit" description changed (wording only, no injection pattern); tool "approve_rule" description changed (wording only, no injection pattern); tool "reject_rule" description changed (wording only, no injection pattern); tool "structural_types" description changed (wording only, no injection pattern); tool "decompose_failure" description changed (wording only, no injection pattern); tool "geometric_confidence" description changed (wording only, no injection pattern); tool "check_realization" description changed (wording only, no injection pattern); tool "check_drift" description changed (wording only, no injection pattern); tool "authorize_execution" description changed (wording only, no injection pattern); tool "load_rule_pack" description changed (wording only, no injection pattern); tool "get_execution_trace" description changed (wording only, no injection pattern); tool "verify_replay" description changed (wording only, no injection pattern); tool "account_status" description changed (wording only, no injection pattern); tool "delete_blueprint" description changed (wording only, no injection pattern); tool "update_blueprint" description changed (wording only, no injection pattern); tool "list_api_keys" description changed (wording only, no injection pattern); tool "rotate_api_key" description changed (wording only, no injection pattern); tool "delete_api_key" description changed (wording only, no injection pattern). Low-severity drift (informational; consistent with the version bump).

• 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)