oh-my-opencode@5.1.8
The Best AI Agent Harness - Batteries-Included OpenCode Plugin with Multi-Model Orchestration, Parallel Background Agents, and Crafted LSP/AST Tools
AI Security Review
scanned 1mo ago · by lpm-firewall-aiSource proves user-invoked agent configuration changes that weaken permission boundaries and enable plugins and hooks. This warrants a capability warning under the supplied policy, without evidence supporting malicious intent or publication blocking.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review
Trigger
User-invoked CLI installation with Codex selected and autonomous permissions enabled, which is the default.
Impact
Subsequent Codex sessions can act with fewer permission boundaries. The inspected permission choice does not itself establish consent to later automatic updates.
Mechanism
The installer updates the user's Codex configuration to disable approval prompts, allow full filesystem access, and enable plugins and trusted hooks.
AI rationale
Source proves user-invoked agent configuration changes that weaken permission boundaries and enable plugins and hooks. This warrants a capability warning under the supplied policy, without evidence supporting malicious intent or publication blocking.
Evidence
dist/cli-node/index.js.codexconfig.toml
Decision evidence
public snapshotAI called this Suspicious at 94.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The Codex configuration helper disables approvals and selects full filesystem access.
- Codex autonomous permissions default to enabled when Codex is selected; an explicit override can disable them.
- The installer targets CODEX_HOME or the user's .codex directory and updates config.toml.
- The CLI installation flow invokes the Codex installer with the selected autonomous permission setting.
- The configuration writer enables plugins, trusts supplied hooks, and persists agent settings.
Evidence against
- The inspected npm postinstall performs package-specific cache cleanup and version checks, rather than invoking the Codex installer.
- The agent integration search hints in both QA scripts are rejection-test inputs; the bundle hints copy bundled MCP runtimes.
- No concrete credential theft or malicious install-time agent configuration mutation was established.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsEvalFilesystem
Source & flagged code
23 flaggedpackage.jsonView file
1Install-time code directly mutates a foreign AI-agent control surface:
L62: "docs/reference/web-terminal-visual-qa.md",
L63: ".opencode/command",
L64: ".opencode/skills",
L65: ".agents/command",
L66: ".agents/skills",
L67: "packages/lsp-tools-mcp/package.json",
...
L97: "packages/prompts-core/prompts/ultrawork/codex.md",
L98: "packages/omo-codex/plugin/.codex-plugin",
L99: "!packages/omo-codex/plugin/node_modules",
...
L147: "test:fast": "bun run script/test-fast.ts",
L148: "test:senpi": "bun run build:senpi-plugin && tsgo --noEmit -p packages/omo-senpi/tsconfig.json && bun test --timeout 20000 packages/omo-senpi && bun test ./.agents/skills/senpi-qa/...
L149: "test:windows-codex": "bun run test:codex",
Write operation from dist/tui.js:
L2183: // packages/utils/src/xdg-data-dir.ts
L2184: i
Critical
Ai Agent Control Hijack
Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.
package.jsonView on unpkg · L1•scripts.postinstall = node postinstall.mjs
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node postinstall.mjs
Findings
4 Critical14 High9 Medium6 Low
CriticalSame File Env Network Executiondist/skills/browser/runtime/omowright/index.js
CriticalDownload Executedist/skills/browser/runtime/omowright/index.js
CriticalRemote Response Code Executiondist/index.js
CriticalAi Agent Control Hijackpackage.json
HighInstall Time Lifecycle Scriptspackage.json
HighChild Processbin/oh-my-opencode.js
HighShelldist/cli-node/index.js
HighEvaldist/skills/ulw-research/scripts/layout-probe.test.ts
HighCredential Redirect Persistencedist/index.js
HighCommand Output Exfiltrationdist/cli-node/index.js
HighSandbox Evasion Gated Capabilitydist/skills/browser/runtime/omowright/index.js
Affected versions and remediation
This report applies to oh-my-opencode@5.1.8.
See all versions of oh-my-opencode for other recorded verdicts.
- Review the evidence and your use of oh-my-opencode@5.1.8 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.