Tachibot MCP Server

In the server's implementation (`dist/src/modes/code-reviewer.js:94`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: *\(/, title: 'Dangerous eval() usage', severity: 'critical' }, { pattern: /document\.write\s*\(/, title:

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/code-reviewer.js

In the server's implementation (`dist/src/modes/commit-guardian.js:386`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: dangerous_eval': 'Avoid eval(). Use safer alternatives', 'dangerous_exec': 'Avoid exec(). Use safer alternat

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/commit-guardian.js

In the server's implementation (`dist/evals/sandbox/python-exec.js:16`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: spawnSync } from 'node:child_process'; import path from 'node:path'; import { fileURLToPath } from 'node:url'; const __

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/evals/sandbox/python-exec.js

In the server's implementation (`dist/evals/writers/jsonl-writer.js:11`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ecFileSync } from 'node:child_process'; export const HARNESS_VERSION = '0.1.0'; function resolveScorerGitSha() { if

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/evals/writers/jsonl-writer.js

In the server's implementation (`dist/src/modes/challenger.js:24`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: port { execSync } from 'child_process'; // import { // renderTable, // renderKeyValueTable, // renderGradientBorde

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/challenger.js

In the server's implementation (`dist/src/modes/commit-guardian.js:387`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: dangerous_exec': 'Avoid exec(). Use safer alternatives', 'xss_risk': 'Sanitize HTML content to prevent XSS',

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/commit-guardian.js

In the server's implementation (`dist/src/modes/scout.js:8`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: port { execSync } from 'child_process'; // import { // renderKeyValueTable, // renderGradientBorderBox, // renderG

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/scout.js

In the server's implementation (`dist/src/tools/claude-integration.js:66`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: const response = await fetch('https://api.anthropic.com/v1/messages', { method: 'POST',

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/tools/claude-integration.js

In the server's implementation (`dist/src/utils/update-checker.js:71`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: const res = await fetch(`https://registry.npmjs.org/${PKG_NAME}`, { signal: controller.signal,

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/utils/update-checker.js

In the server's implementation (`dist/src/modes/challenger.js:85`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: const dateStr = execSync('date "+%B %d, %Y"', { encoding: 'utf8' }).trim(); return dateStr; } ca

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/challenger.js

In the server's implementation (`dist/src/modes/scout.js:41`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: const dateStr = execSync('date "+%B %d, %Y"', { encoding: 'utf8' }).trim(); return dateStr; } ca

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/modes/scout.js

In the server's implementation (`dist/src/memory/memory-manager.js:191`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ories */ async import(items) { if (!this.provider) { throw new Error('Memory manager not in

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/memory/memory-manager.js

In the server's implementation (`dist/src/memory/providers/devlog-provider.js:213`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: } } async import(items) { let imported = 0; for (const item of items) { try {

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/memory/providers/devlog-provider.js

In the server's implementation (`dist/src/memory/providers/dokoro-provider.js:185`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ues()); } async import(items) { let imported = 0; for (const item of items) { try {

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/memory/providers/dokoro-provider.js

In the server's implementation (`dist/src/memory/providers/hybrid-provider.js:288`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ues()); } async import(items) { let totalImported = 0; // Import to both providers if (t

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/memory/providers/hybrid-provider.js

In the server's implementation (`dist/src/memory/providers/local-provider.js:234`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ues()); } async import(items) { let imported = 0; for (const item of items) { try {

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/memory/providers/local-provider.js

In the server's implementation (`dist/src/memory/providers/mem0-provider.js:256`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: } } async import(items) { if (!this.isConnected) { throw new Error('Mem0 provider not

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/src/memory/providers/mem0-provider.js

"tachibot-mcp" executes postinstall script(s) at install time. An install hook runs at install time; most are routine build/setup, but review what it does before trusting it.

Evidence: node scripts/postinstall-hint.js || true

Fix: Review the scripts; install with --ignore-scripts where possible and vet what they do.

Location: package tachibot-mcp