io.github.lastmanupinc-hub/axis-toolbox

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).

  tool "iliad_web_research" description: cross-tool redirection ('instead of call…'). tool "iliad_web_research_crawl" description: cross-tool redirection ('instead of call…'). tool "axis_validate" description: off-vendor URL in a description (api.avatar.jonathanarvay.com); capability text should not name where data goes. tool "axis_inspect" description: off-vendor URL in a description (api.avatar.jonathanarvay.com); capability text should not name where data goes. tool "axis_compare" description: off-vendor URL in a description (api.avatar.jonathanarvay.com); capability text should not name where data goes. tool "axis_manifest_verify" description: off-vendor URL in a description (api.avatar.jonathanarvay.com); capability text should not name where data goes. tool "roblox_compliance_check" description: off-vendor URL in a description (api.avatar.jonathanarvay.com); capability text should not name where data goes.

• 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 "prepare_agentic_purchasing" has a secret-shaped input parameter "referral_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 0.4.0 (minor bump): tool "analyze_repo" description changed (wording only, no injection pattern); tool "analyze_files" description changed (wording only, no injection pattern); tool "list_programs" description changed (wording only, no injection pattern); tool "get_snapshot" description changed (wording only, no injection pattern); tool "prepare_agentic_purchasing" description changed (wording only, no injection pattern); tool "closer" description changed (wording only, no injection pattern); tool "deploy" description changed (wording only, no injection pattern); tool "search_and_discover_tools" description changed (wording only, no injection pattern); tool "ping_payment" description changed (wording only, no injection pattern); tool "improve_my_agent_with_axis" description changed (wording only, no injection pattern); tool "iliad_web_research" description changed (wording only, no injection pattern); tool "iliad_web_research_crawl" description changed (wording only, no injection pattern); tool "iliad_object_storage" description changed (wording only, no injection pattern); tool "iliad_vector_database" description changed (wording only, no injection pattern); tool "iliad_document_parsing" description changed (wording only, no injection pattern); tool "iliad_text_to_speech" description changed (wording only, no injection pattern); tool "iliad_speech_to_text" description changed (wording only, no injection pattern); tool "assemble_representment" description changed (wording only, no injection pattern); new tool "delete_snapshot" touches a sensitive surface (email/exec/net/fs); new tool "get_install_manifest" touches a sensitive surface (email/exec/net/fs); new tool "discover_estate_tools" appeared; new tool "axis_validate" whose description contains embedded outbound URL; new tool "axis_inspect" whose description contains embedded outbound URL; new tool "axis_compare" whose description contains embedded outbound URL; new tool "axis_manifest_verify" whose description contains embedded outbound URL; new tool "roblox_compliance_check" whose description contains embedded outbound URL; new capability primitive(s) in source: http-client; new hardcoded outbound host/BCC destination(s): axis-api-6c7z.onrender.com, fast.io, firecrawl.dev, github.com, replicate.com, resend.com.

• Transport config

  pass10% of grade

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

  remote endpoint "https://axis-api-6c7z.onrender.com/mcp" uses TLS (https); remote endpoint uses OAuth (strongest auth mode)