com.decision-anchor/da

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

  pass25% of grade

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

  Scanned 312 text field(s); no instruction-injection patterns found. PASS = no static red flag. A server can serve different tool descriptions at runtime (the rug-pull class), which static scanning cannot see.

• Supply chain

  pass25% of grade

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

  repository linkage verified

• Credential hygiene

  fail15% of grade

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

  tool "get_agent_profile" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_decision" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "confirm_decision" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "propose_bilateral" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_decision" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "list_decisions" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "observe_environment" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "observe_pattern" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "register_tool" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "purchase_tool" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_ise_session" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_ise_status" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "exit_ise_session" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_dac_balance" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_dac_ur" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_trial_status" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_sdac_session" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "run_sdac_trial" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_sdac_session" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "end_sdac_session" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "list_classifications" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "compare_anomaly" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_evidence_report" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_environment_anomaly" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_decision_metadata_distribution" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential). tool "get_self_classification_distribution" has a secret-shaped input parameter "auth_token" — secrets must never be tool parameters (the model would supply/handle the credential).

• Permission scope

  pass15% of grade

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

  No shell-exec, unbounded/sensitive filesystem, or download-and-run primitives found in scanned source. PASS = no static red flag. Static scanning is blind to code fetched or executed at runtime.

• 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 1.3.36 (patch bump): tool "get_agent_profile" description changed (wording only, no injection pattern); tool "create_decision" description changed (wording only, no injection pattern); tool "create_decision" gained input property(ies): decision_at; tool "confirm_decision" description changed (wording only, no injection pattern); tool "propose_bilateral" description changed (wording only, no injection pattern); tool "propose_bilateral" gained input property(ies): decision_at; tool "get_decision" description changed (wording only, no injection pattern); tool "observe_environment" description changed (wording only, no injection pattern); tool "observe_pattern" description changed (wording only, no injection pattern); tool "purchase_tool" description changed (wording only, no injection pattern); tool "create_ise_session" description changed (wording only, no injection pattern); tool "get_ise_status" description changed (wording only, no injection pattern); tool "exit_ise_session" description changed (wording only, no injection pattern); tool "get_dac_balance" description changed (wording only, no injection pattern); tool "get_dac_ur" description changed (wording only, no injection pattern); tool "get_trial_status" description changed (wording only, no injection pattern); tool "create_sdac_session" description changed (wording only, no injection pattern); tool "run_sdac_trial" description changed (wording only, no injection pattern); tool "get_sdac_session" description changed (wording only, no injection pattern); tool "end_sdac_session" description changed (wording only, no injection pattern); tool "get_evidence_report" description changed (wording only, no injection pattern); tool "get_environment_anomaly" description changed (wording only, no injection pattern); tool "get_decision_metadata_distribution" description changed (wording only, no injection pattern); tool "get_decision_metadata_distribution" gained input property(ies): payment_signature; tool "submit_feedback" description changed (wording only, no injection pattern); new hardcoded outbound host/BCC destination(s): github.com.

• 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)