registry / @inharness-ai/agent-adapters / 0.8.6

@inharness-ai/agent-adapters@0.8.6

Unified TypeScript interface for AI agent SDKs — Claude Code, Codex, OpenCode, Gemini

AI Security Review

scanned 1d ago · by lpm-firewall-ai

No confirmed malicious attack surface. Adapter execution is explicit and forwards caller-provided prompts, credentials, MCP settings, images, and workspace configuration to selected agent SDKs.

Static reason

No blocking static signals were detected.

Trigger

Explicit application call to an adapter's execute() method.

Impact

Expected agent capabilities may act in the caller-selected workspace; no covert persistence, harvesting, or exfiltration was found.

Mechanism

User-configured AI-agent SDK orchestration with temporary skill/image staging.

Rationale

The package is a user-invoked adapter layer for Claude, Codex, OpenCode, and Gemini SDKs. Its filesystem, network, shell, and credential-related behavior is scoped to explicit adapter execution and caller-provided configuration, with no install-time execution or concrete malicious chain.

Evidence

package.jsondist/chunk-75JOMN6X.cjsdist/chunk-Y4FUFQLA.cjsdist/chunk-2RH7DQ5V.cjsdist/chunk-Z3RJOCLW.cjsdist/testing/index.cjs

Decision evidence

public snapshot

AI called this Clean at 95.0% confidence as Benign with low false-positive risk.

Evidence for block

@inharness-ai/agent-adapters@0.8.6

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code