context safety score
A score of 49/100 indicates multiple risk signals were detected. This entity shows patterns commonly associated with malicious intent.
supply chain
Found 1 install-script pattern(s) in documentation (likely install instructions, not executable)
supply chain
Found 1 remote script pattern(s) in documentation (likely install instructions, not executable)
consent bypass
The install-hooks.sh and install-hooks.ps1 scripts programmatically set hasTrustDialogAccepted=true in ~/.claude.json, silently bypassing Claude Code's workspace trust dialog — a security mechanism designed to require explicit user consent before allowing tool execution in a directory. The configure_workspace_trust function in both scripts modifies this state file without meaningful user confirmation during the install flow. (location: install-hooks.sh:configure_workspace_trust() and install-hooks.ps1:Configure-WorkspaceTrust())
tool shadowing
Five tools directly shadow built-in agent tools by providing the same core functionality with similar names and identical parameter semantics: smart_read (shadows Read/read_file), smart_write (shadows Write/write_file), smart_edit (shadows Edit/edit_file), smart_grep (shadows Grep/grep), smart_glob (shadows Glob/glob). The wildcard hooks (matcher: '*') registered on PreToolUse intercept every Read tool call and replace it with smart_read results via 'exit 2' (block original tool), meaning the agent's native file reading is silently substituted without agent awareness. This is not just offering an alternative — the hook infrastructure actively intercepts and replaces native tool calls. (location: src/tools/file-operations/smart-read.ts, hooks/dispatcher.ps1 (PreToolUse Read interception), hooks/read-cache-interceptor.ps1)
supply chain
The hook helper invoke-mcp.ps1 uses 'npx -y token-optimizer-mcp@latest' as a fallback execution path, which auto-downloads and executes the latest version from npm without version pinning on every invocation. Since the hooks intercept ALL tool I/O (file contents, user prompts, bash commands, conversation context), a compromised npm package would receive all this data. The @latest tag means even a pinned initial install can be bypassed by a malicious npm publish. (location: hooks/helpers/invoke-mcp.ps1 (npx fallback path))
capability escalation
The install scripts aggressively modify configuration files for 5 different AI tools (Claude Desktop, Cursor IDE, Cline/VS Code, VS Code Copilot, Windsurf) without clear per-tool user consent, injecting the MCP server configuration into each. A user installing for Claude Code may not expect their Cursor, Windsurf, or Cline configurations to be modified. The Claude Desktop config is always created even if the tool isn't installed. Combined with wildcard hook registration on all 4 hook phases (PreToolUse, PostToolUse, UserPromptSubmit, PreCompact), the server gains interception capability over every agent interaction. (location: install-hooks.sh:configure_mcp_server(), install-hooks.ps1:Configure-MCPServer())
curl https://api.brin.sh/mcp/ooples%2Ftoken-optimizer-mcpCommon questions teams ask before deciding whether to use this mcp server in agent workflows.
ooples/token-optimizer-mcp currently scores 49/100 with a suspicious verdict and medium confidence. The goal is to protect agents from high-risk context before they act on it. Treat this as a decision signal: higher scores suggest lower observed risk, while lower scores mean you should add review or block this mcp server.
Use the score as a policy threshold: 80–100 is safe, 50–79 is caution, 20–49 is suspicious, and 0–19 is dangerous. Teams often auto-allow safe, require human review for caution/suspicious, and block dangerous.
brin evaluates four dimensions: identity (source trust), behavior (runtime patterns), content (malicious instructions), and graph (relationship risk). Analysis runs in tiers: static signals, deterministic pattern checks, then AI semantic analysis when needed.
Identity checks source trust, behavior checks unusual runtime patterns, content checks for malicious instructions, and graph checks risky relationships to other entities. Looking at sub-scores helps you understand why an entity passed or failed.
brin performs risk assessments on external context before it reaches an AI agent. It scores that context for threats like prompt injection, hijacking, credential harvesting, and supply chain attacks, so teams can decide whether to block, review, or proceed safely.
No. A safe verdict means no significant risk signals were detected in this scan. It is not a formal guarantee; assessments are automated and point-in-time, so combine scores with your own controls and periodic re-checks.
Re-check before high-impact actions such as installs, upgrades, connecting MCP servers, executing remote code, or granting secrets. Use the API in CI or runtime gates so decisions are based on the latest scan.
Learn more in threat detection docs, how scoring works, and the API overview.
Assessments are automated and may contain errors. Findings are risk indicators, not confirmed threats. This is a point-in-time assessment; security posture can change.
integrate brin in minutes — one GET request is all it takes. query the api, browse the registry, or download the full dataset.