agent-cards@0.6.19
CLI for managing virtual debit cards for AI agents
AI Security Review
scanned 3mo ago · by lpm-firewall-aiNo concrete malicious install-time or covert exfiltration behavior was found. The package nevertheless has a real persistent AI-agent configuration and self-update capability, so it should be warned rather than marked clean.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
Explicit `agent-cards agents add` followed by later CLI execution; an outdated global CLI invocation for self-update.
Impact
Can persist package-authored behavioral guidance in agent control surfaces and replace the globally installed CLI.
Mechanism
Global AI-agent instruction-file mutation and runtime npm self-update.
AI rationale
No concrete malicious install-time or covert exfiltration behavior was found. The package nevertheless has a real persistent AI-agent configuration and self-update capability, so it should be warned rather than marked clean.
Evidence
package.jsondist/bin/agent-cards.jsdist/src/commands/agents.jsdist/src/lib/update.jsdist/src/lib/global-install.jsdist/src/commands/setup-mcp.js
Network endpoints4
registry.npmjs.org/agent-cards/latestt.agentcard.shapi.agentcard.shmcp.agentcard.sh/mcp
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- dist/src/commands/agents.js writes managed blocks to global Claude, Codex, and Gemini instruction files.
- dist/bin/agent-cards.js refreshes existing managed agent blocks on ordinary CLI runs.
- dist/src/lib/update.js can fetch npm metadata and run global package-manager self-updates at runtime.
Evidence against
- package.json has no preinstall, install, or postinstall hook; only prepublishOnly.
- Agent configuration is initiated by the explicit `agents add` command and refreshes only package-marked blocks.
- Network and credential handling are aligned with the card-service CLI; no secret harvesting or unrelated exfiltration was found.
- Shell use is limited to package-manager update/install and repository detection; no arbitrary remote payload execution found.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShell
Source & flagged code
4 flaggeddist/src/lib/update.jsView file
2import { join, dirname } from 'path';
L3: import { execSync, spawnSync, execFileSync } from 'child_process';
L4: import { CONFIG_DIR } from './config.js';
High
Child Process
Package source references child process execution.
dist/src/lib/update.jsView on unpkg · L2dist/src/commands/api.jsView file
•matchType = previous_version_dangerous_delta
matchedPackage = agent-cards@0.5.66
matchedIdentity = npm:YWdlbnQtY2FyZHM:0.5.66
similarity = 0.463
summary = stored previous version shares package body but lacks this dangerous source file
Critical
Previous Version Dangerous Delta
This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
dist/src/commands/api.jsView on unpkg22if (existsSync(p)) {
L23: return (await import(pathToFileURL(p).href));
L24: }
Findings
1 Critical3 High4 Medium7 Low
CriticalPrevious Version Dangerous Deltadist/src/commands/api.js
HighChild Processdist/src/lib/update.js
HighShell
HighRuntime Package Installdist/src/lib/global-install.js
MediumDynamic Requiredist/src/commands/api.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowNon Install Lifecycle Scripts
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowTelemetry
LowUrl Strings
LowNo License
Affected versions and remediation
This report applies to agent-cards@0.6.19.
See all versions of agent-cards for other recorded verdicts.
- Review the evidence and your use of agent-cards@0.6.19 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.