io.github.beepboop2025/seiche

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 (seiche.info); capability text should not name where data goes. server instructions: off-vendor URL in a description (t.me); capability text should not name where data goes. server instructions: off-vendor URL in a description (api.liquilens.in); capability text should not name where data goes. server instructions: off-vendor URL in a description (groundcheck.seiche.info.); capability text should not name where data goes. server instructions: off-vendor URL in a description (api.seiche.info); 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

  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. seiche/accounts.py) — 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 (seiche/alerts.py: subprocess.* call; seiche/ai.py: outbound HTTP/socket call). has both shell/exec and destructive filesystem writes (seiche/alerts.py: subprocess.* call; seiche/accounts.py: destructive filesystem call (shutil.rmtree/os.remove/etc.)). discloses shell/exec capability: seiche/alerts.py: subprocess.* call discloses broad/destructive filesystem access: seiche/accounts.py: destructive filesystem call (shutil.rmtree/os.remove/etc.) 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 0.10.1 (minor bump): tool "world_markets_context" description changed (wording only, no injection pattern); new capability primitive(s) in source: http-client, subprocess; new hardcoded outbound host/BCC destination(s): alice.btc.calendar.opentimestamps.org, api.exchange.coinbase.com, api.fiscaldata.treasury.gov, api.seiche.info, api.telegram.org, bob.btc.calendar.opentimestamps.org, data-api.ecb.europa.eu, data.financialresearch.gov, finney.calendar.eternitywall.com, fonts.googleapis.com, fonts.gstatic.com, fred.stlouisfed.org, github.com, markets.newyorkfed.org, palimpsest.info, publicreporting.cftc.gov, raw.githubusercontent.com, schema.org, seiche.info, stablecoins.llama.fi, stats.bis.org, www.federalreserve.gov, www.sitemaps.org, www.w3.org, x.y, x402.org.

• 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); sends telemetry to third-party SDK (amplitude) — disclosure, verify opt-in; manifest/source references analytics behavior — disclosure