@seoagent-official/seoagent@1.100.8
The persistent AI SEO agent for Claude Code. Audits, keyword strategy, briefs, articles, real product screenshots from your repo, and the autopilot loop (cloud detects → CLI executes → ack closes) — other SEO tools write the prompt, SEOAgent runs it.
AI Security Review
scanned 17d ago · by lpm-firewall-aiThe source confirms that cloud inbox actions can drive shell commands and project edits through Claude with automatic edit approval. Because this capability is exposed through an explicitly invoked CLI command and no install-time trigger or concrete attack behavior was found, warn rather than block.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
The user runs `seoagent process`; `--yes` also skips action selection and processes every pending item.
Impact
A processed action can cause project changes or shell commands to run in the current working directory. The source supports a conditional capability risk, not a confirmed malicious action.
Mechanism
The CLI fetches authenticated inbox actions, turns each action into instructions for the Claude Agent SDK, and grants it Bash and file-editing tools with automatic edit approval.
AI rationale
The source confirms that cloud inbox actions can drive shell commands and project edits through Claude with automatic edit approval. Because this capability is exposed through an explicitly invoked CLI command and no install-time trigger or concrete attack behavior was found, warn rather than block.
Evidence
index.js.seoagent/inbox/project files in the current working directory
Network endpoints1
seoagent.com/api/cli/actions/fetch
Decision evidence
public snapshotAI called this Suspicious at 88.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The package exposes a `seoagent` CLI; its description identifies a cloud-to-CLI autopilot workflow.
- The `process` command syncs the inbox and lets the user choose actions; `--yes` skips those prompts and processes every pending action.
- The CLI fetches pending actions from its configured API base using an authorization header.
- The agent prompt tells Claude to apply each inbox action to the current project’s codebase.
- The Claude Agent SDK receives Bash, Write, and Edit tools with automatic edit approval.
- The `.claude/settings.json` integration adds SEOAgent sync hooks after a skill refresh; the hook runs the package’s sync command.
Evidence against
- The complete manifest has no npm lifecycle hook; agent processing is exposed through an explicit CLI command.
- The inspected agent configuration behavior is for package sync, not a broad agent control-surface change.
Behavioral surface
EnvironmentVarsNetwork
HighEntropyStrings
Source & flagged code
7 flaggedindex.jsView file
85- Bundle scaffolded.
L86: `,"utf-8"),r.push("log.md")),{created:r,okfDir:n}}import{log as At}from"@clack/prompts";var u={info:e=>At.info(e),warn:e=>At.warn(e),error:e=>At.error(e),success:e=>At.success(e),s...
L87: `))}function Kf(){return`The setup code did not bind this project. Do not continue without a successful login \u2014 get a fresh code at seoagent.com/grok, then run \`${h("login")}...
Critical
Same File Env Network Execution
A single source file combines environment access, network access, and code or shell execution with blocking evidence.
index.jsView on unpkg · L8585Trigger-reachable chain: manifest.main -> index.js
L85: - Bundle scaffolded.
L86: `,"utf-8"),r.push("log.md")),{created:r,okfDir:n}}import{log as At}from"@clack/prompts";var u={info:e=>At.info(e),warn:e=>At.warn(e),error:e=>At.error(e),success:e=>At.success(e),s...
L87: `))}function Kf(){return`The setup code did not bind this project. Do not continue without a successful login \u2014 get a fresh code at seoagent.com/grok, then run \`${h("login")}...
Critical
Trigger Reachable Dangerous Capability
A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.
index.jsView on unpkg · L8585- Bundle scaffolded.
L86: `,"utf-8"),r.push("log.md")),{created:r,okfDir:n}}import{log as At}from"@clack/prompts";var u={info:e=>At.info(e),warn:e=>At.warn(e),error:e=>At.error(e),success:e=>At.success(e),s...
L87: `))}function Kf(){return`The setup code did not bind this project. Do not continue without a successful login \u2014 get a fresh code at seoagent.com/grok, then run \`${h("login")}...
Findings
2 Critical3 High5 Medium2 Low
CriticalSame File Env Network Executionindex.js
CriticalTrigger Reachable Dangerous Capabilityindex.js
HighCommand Output Exfiltrationindex.js
HighTrigger Reachable Command Output Exfiltrationindex.js
HighSemantic Analysis Limitedindex.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
MediumAi Review Evidencepackage.json
MediumAi Review Evidenceindex.js
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @seoagent-official/seoagent@1.100.8.
See all versions of @seoagent-official/seoagent for other recorded verdicts.
- Review the evidence and your use of @seoagent-official/seoagent@1.100.8 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.