io.github.desplega-ai/agent-swarm

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

  unscannable25% of grade

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

  No tool descriptions, server instructions, prompts, or resources were fetched; nothing to scan for injection.

• Supply chain

  warn25% of grade

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

  dependency "ai" is edit-distance 2 from popular package "chai" (typosquat candidate); dependency "e2b" is edit-distance 2 from popular package "ejs" (typosquat candidate); dependency "ink" is edit-distance 2 from popular package "pino" (typosquat candidate); dependency "yjs" is edit-distance 1 from popular package "ejs" (typosquat candidate); dependency "hono" is edit-distance 2 from popular package "pino" (typosquat candidate); dependency "viem" is edit-distance 2 from top-20 package "vue" (typosquat candidate); dependency "@inkjs/ui" is edit-distance 2 from top-20 package "vue" (typosquat candidate); dependency "@x402/evm" is edit-distance 2 from popular package "ejs" (typosquat candidate); dependency "@x402/core" is edit-distance 1 from popular package "cors" (typosquat candidate); dependency "@openai/codex" is edit-distance 2 from popular package "@babel/core" (typosquat candidate); dependency "@opentelemetry/api" is edit-distance 2 from popular package "pg" (typosquat candidate); dependency "@earendil-works/pi-ai" is edit-distance 2 from popular package "chai" (typosquat candidate)

• Credential hygiene

  pass15% of grade

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

  No hardcoded secrets, secret-shaped tool parameters, or leaked keys found. PASS = no static red flag; static analysis cannot prove the code honors env-based secret handling at runtime.

• 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 1.143.0 (minor bump): new hardcoded outbound host/BCC destination(s): 127.0.0.1, agent-swarm.dev, api.desplega.agent-swarm.dev, api.kapso.ai, api.slack.com, api.swarm.localhost, developer.atlassian.com, docs.agent-swarm.dev, github.com, linear.app, openrouter.ai, ui.swarm.localhost, your-dashboard.com.

• Transport config

  warn10% of grade

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

  manifest/source references tracking behavior — disclosure