registry / @inharness-ai/agent-adapters / 0.8.5

@inharness-ai/agent-adapters@0.8.5

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

AI Security Review

scanned 11h ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. A user-invoked OpenCode adapter run can materialize caller-supplied skills and mirror them into the current project's `.opencode/skills` directory. The package removes the generated UUID-prefixed directories during cleanup; no install-time attack surface is established.

Static reason

No blocking static signals were detected.

Trigger

Calling `OpencodeAdapter.execute()` with `params.skills`.

Impact

Caller-provided skill content can be exposed to an OpenCode agent with configured edit/bash permissions during that run.

Mechanism

Scoped runtime AI-agent skill materialization and OpenCode execution.

Rationale

No concrete malicious behavior was found after source inspection, but the package intentionally creates a project-local AI-agent extension and launches user-invoked agents with broad tool permissions. Per the lifecycle policy, this merits a warning rather than a block.

Evidence

package.jsondist/chunk-OEPE3RNC.jsdist/chunk-TYWZAXEU.jsdist/chunk-SHUY6KIA.jsdist/testing/index.js.opencode/skills/agent-adapters-<uuid>-<slug>/SKILL.md

Decision evidence

public snapshot

AI called this Suspicious at 88.0% confidence as Dangerous Capability with low false-positive risk.

@inharness-ai/agent-adapters@0.8.5

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code