io.github.human-pages-ai/humanpages

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 "register_agent" param "accept_tos" description: imperative telling the model it must read/send/attach something. tool "register_agent" param "accept_tos" description: off-vendor URL in a description (humanpages.ai); capability text should not name where data goes. tool "set_wallet" description: bare 'IMPORTANT:' prose in a description (tag form is high-confidence; prose is warn). tool "set_wallet" description: 'before using this tool…' preamble (tool-poisoning shape). tool "create_job_offer" description: bare 'IMPORTANT:' prose in a description (tag form is high-confidence; prose is warn). tool "create_job_offer" description: 'before using this tool…' preamble (tool-poisoning shape). tool "mark_job_paid" description: bare 'IMPORTANT:' prose in a description (tag form is high-confidence; prose is warn). tool "mark_job_paid" description: 'before using this tool…' preamble (tool-poisoning shape). tool "approve_completion" description: bare 'IMPORTANT:' prose in a description (tag form is high-confidence; prose is warn). tool "start_stream" description: bare 'IMPORTANT:' prose in a description (tag form is high-confidence; prose is warn). tool "start_stream" description: imperative telling the model it must read/send/attach something. tool "pause_stream" description: imperative telling the model it must read/send/attach something.

• 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 "create_job_offer" has a secret-shaped input parameter "callback_secret" — secrets must never be tool parameters (the model would supply/handle the credential). tool "create_listing" has a secret-shaped input parameter "callback_secret" — 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

  pass10% of grade

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

  tool surface, capabilities, install scripts, and publisher unchanged vs prior version 1.2.3 (minor bump); no rug-pull drift.

• 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)