io.github.philpof102-svg/biii

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

  fail25% of grade

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

  tool "till_key_exposure" description: names an outbound copy/exfil destination (BCC/exfil verb).

• Supply chain

  pass25% of grade

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

  repository linkage verified

• Credential hygiene

  pass15% of grade

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

  Reads secrets from the environment (e.g. bin/biii-mcp.js) — the recommended intake shape. PASS = no static red flag; static analysis cannot prove the code honors env-based secret handling at runtime.

• Permission scope

  warn15% of grade

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

  has both shell/exec and network egress — capable of download-and-run, though no such path was observed (lib/agent-vet.js: imports child_process; lib/agent-vet.js: outbound HTTP/socket call). has both shell/exec and destructive filesystem writes (lib/agent-vet.js: imports child_process; lib/usage.js: destructive/writing fs.* call (write/unlink/rm/chmod/rename)). discloses shell/exec capability: lib/agent-vet.js: imports child_process discloses broad/destructive filesystem access: lib/usage.js: destructive/writing fs.* call (write/unlink/rm/chmod/rename) discloses broad/destructive filesystem access: lib/wallet-watch.js: destructive/writing fs.* call (write/unlink/rm/chmod/rename) discloses broad/destructive filesystem access: lib/x402-settle.js: destructive/writing fs.* call (write/unlink/rm/chmod/rename) Capability DISCLOSURE, not a verdict: static analysis sees the primitive is present and reachable, not whether its use is attacker-controlled.

• 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 0.1.0 (minor bump): tool "till_trust" gained input property(ies): rail; tool "till_create_invoice" gained input property(ies): rail; tool "till_check_invoice" description changed (wording only, no injection pattern); tool "till_check_invoice" gained input property(ies): rail; tool "till_vet_meme" description changed (wording only, no injection pattern); tool "till_vet_meme" gained input property(ies): siblingCount; tool "till_export" description changed (wording only, no injection pattern); tool "till_meter" description changed (wording only, no injection pattern); tool "till_funder_history" description changed (wording only, no injection pattern); new capability primitive(s) in source: child_process, fs-write, http-client; new hardcoded outbound host/BCC destination(s): 127.0.0.1, api.dexscreener.com, api.gopluslabs.io, api.honeypot.is, api.trongrid.io, arbitrum.blockscout.com, avisradar-production.up.railway.app, base.blockscout.com, basescan.org, beaconlayer.co, dexscreener.com, eth.blockscout.com, ethereum-rpc.publicnode.com, gnosis.blockscout.com, mainnet.base.org, optimism.blockscout.com, polygon.blockscout.com, raw.githubusercontent.com, robinhoodchain.blockscout.com, x, your-biii-node.

• 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)