@respira/wordpress-mcp-server@8.3.7: AI-agent capability abuse

AI Security Review

scanned 2mo ago · by lpm-firewall-ai

This is not malicious: the observed agent configuration changes are explicit user-invoked setup features, not unconsented install-time behavior. It merits a warning because the commands can alter AI-agent behavior and one path accepts a remote configuration payload.

Static reason

No blocking static signals were detected.

Trigger

The user runs install-skills or --install-config with a selected client and install token.

Impact

It can add Respira-controlled instructions or an MCP server entry to a chosen agent environment.

Mechanism

Explicit AI-agent skill and MCP configuration installation.

AI rationale

This is not malicious: the observed agent configuration changes are explicit user-invoked setup features, not unconsented install-time behavior. It merits a warning because the commands can alter AI-agent behavior and one path accepts a remote configuration payload.

Evidence

package.jsondist/install-skills.jsdist/setup.js~/.claude/skills/<skill>/SKILL.md~/.codex/config.toml~/.claude.json.cursor/mcp.json~/.antigravity/mcp.json

Network endpoints1

www.respira.press

Decision evidence

public snapshot

AI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.

Evidence for warning

An explicit install-skills command copies bundled skill files into the Claude Code skills directory.
An explicit install-config command fetches a Respira configuration payload and writes one named MCP server entry into selected AI-client config files.

Evidence against

package.json has no preinstall, install, or postinstall lifecycle hook.
AI-agent configuration and skill writes require user-supplied command-line flags or an install command; no automatic installation path was found.
The config merge replaces only the selected server name and preserves other configured servers.

Behavioral surface

Source

ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStringsPossibleObfuscationUrlStrings

ManifestNo manifest risk signals triggered.

scanned 23 file(s), 1021 KB of source, external domains: acme.test, example.com, nodejs.org, registry.npmjs.org, respira.press, www.respira.press, yoursite.com

Source & flagged code

6 flagged

dist/server.jsView file

•Manifest-reachable source resolves another installed package, overwrites its runtime code, and injects package-defined remote behavior. dist/server.js: import { existsSync, mkdirSync, readFileSync, writeFileSync } from 'fs'; writeFileSync(STATE_FILE_PATH, JSON.stringify(next, null, 2), { encoding: 'utf8', mode: 0o600 }); Follow these on every site. They are what separates a great result from a frustrating one, and most are one extra step, not one extra tool. The Respira MCP is a single-person project actively under development. Real customers hit real bugs, and the fastest fix path is a structured report from the agent that watched it ... 5. **User says yes** → call \`respira_report_issue\` with a clear title (the symptom), brief (one paragraph of what happened), severity (blocker if the w

High

Entrypoint Foreign Package Code Overwrite

Manifest-reachable source overwrites another installed package with package-defined remote behavior.

dist/server.jsView on unpkg

dist/install-skills.jsView file

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. const bundled_skills_dir = path.resolve(__dirname, '..', 'skills'); const c = { reset: '\x1b[0m', bold: '\x1b[1m', dim: '\x1b[2m', red: '\x1b[31m', green: '\x1b[32m', yellow: '\x1b[33m', blue: '\x1b[34m', cyan: '\x1b[36m', }; function parseargs(argv) { const opts = { dest: path.join(os.homedir(), '.claude', 'skills'), force: false, help: false, }; for (let i = 0; i < argv.length; i++) { const arg = argv[i]; if (arg === '--help' || arg === '-h') { opts.help = true; } else if (arg === '--force' || arg === '-f') { opts.force = true; } else if (

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

dist/install-skills.jsView on unpkg

38function parseArgs(argv) { const opts = { dest: path.join(os.homedir(), '.claude', 'skills'), force: false, help: false,

Medium

Ai Review Evidence

An explicit install-skills command copies bundled skill files into the Claude Code skills directory.

dist/install-skills.jsView on unpkg · L38

114function copySkill(slug, srcDir, destRoot, force) { const destDir = path.join(destRoot, slug); const destFile = path.join(destDir, 'SKILL.md'); const exists = fs.existsSync(destFile); if (exists && !force) { return 'skipped'; } fs.mkdirSync(destDir, { recursive: true }); copyFile(path.join(srcDir, slug, 'SKILL.md'), destFile);

Medium

Ai Review Evidence

An explicit install-skills command copies bundled skill files into the Claude Code skills directory.

dist/install-skills.jsView on unpkg · L114

dist/setup.jsView file

651async function fetchInstallConfig(client, installToken) { const endpoint = `${RESPIRA_WEBSITE_URL.replace(/\/$/, '')}/api/mcp/config/export-by-token?token=${encodeURIComponent(installToken)}`; const response = await fetch(endpoint, { headers: { ...getRespiraAgentHeaders(), 'X-Respira-MCP-Version': 'install-config', }, }); const data = (await response.json().catch(() => ({}))); if (!response.ok) { throw new Error(data?.error || `Failed to fetch install config (${response.status})`); } if (data?.client !== client) {

Medium

Ai Review Evidence

An explicit install-config command fetches a Respira configuration payload and writes one named MCP server entry into selected AI-client config files.

dist/setup.jsView on unpkg · L651

704async function installClientConfig(client, installToken) { print(`Installing Respira MCP config for ${client}...`, c.cyan); const payload = await fetchInstallConfig(client, installToken); const targetPath = getTargetConfigPath(client); const existingRaw = fs.existsSync(targetPath) ? fs.readFileSync(targetPath, 'utf8') : null; const merged = client === 'codex' ? mergeTomlServerConfig(existingRaw, payload.content, payload.serverName) : mergeJsonServerConfig(existingRaw, payload.content, payload.serverName); ensureParentDirectory(targetPath); fs.write

Medium

Ai Review Evidence

An explicit install-config command fetches a Respira configuration payload and writes one named MCP server entry into selected AI-client config files.

dist/setup.jsView on unpkg · L704

Source & flagged code

6 flagged · loading source

dist/server.jsView file

•Manifest-reachable source resolves another installed package, overwrites its runtime code, and injects package-defined remote behavior. dist/server.js: import { existsSync, mkdirSync, readFileSync, writeFileSync } from 'fs'; writeFileSync(STATE_FILE_PATH, JSON.stringify(next, null, 2), { encoding: 'utf8', mode: 0o600 }); Follow these on every site. They are what separates a great result from a frustrating one, and most are one extra step, not one extra tool. The Respira MCP is a single-person project actively under development. Real customers hit real bugs, and the fastest fix path is a structured report from the agent that watched it ... 5. **User says yes** → call \`respira_report_issue\` with a clear title (the symptom), brief (one paragraph of what happened), severity (blocker if the w

High

Entrypoint Foreign Package Code Overwrite

Manifest-reachable source overwrites another installed package with package-defined remote behavior.

dist/server.jsView on unpkg

dist/install-skills.jsView file

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. const bundled_skills_dir = path.resolve(__dirname, '..', 'skills'); const c = { reset: '\x1b[0m', bold: '\x1b[1m', dim: '\x1b[2m', red: '\x1b[31m', green: '\x1b[32m', yellow: '\x1b[33m', blue: '\x1b[34m', cyan: '\x1b[36m', }; function parseargs(argv) { const opts = { dest: path.join(os.homedir(), '.claude', 'skills'), force: false, help: false, }; for (let i = 0; i < argv.length; i++) { const arg = argv[i]; if (arg === '--help' || arg === '-h') { opts.help = true; } else if (arg === '--force' || arg === '-f') { opts.force = true; } else if (

Findings

2 High6 Medium4 Low

HighEntrypoint Foreign Package Code Overwritedist/server.js

HighTrigger Reachable External Ai Agent Control Surface Mutationdist/install-skills.js

MediumNetwork

MediumEnvironment Vars

MediumAi Review Evidencedist/install-skills.js

MediumAi Review Evidencedist/setup.js

LowScripts Present

LowFilesystem

LowHigh Entropy Strings

LowUrl Strings

@respira/wordpress-mcp-server@8.3.7

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Runtime surface

Behavioral surface

Source & flagged code