- CLI: avoid loading provider discovery during startup model normalization. (#46522) Thanks @ItsAditya-xyz and @vincentkoc.
- Agents/Telegram: avoid rebuilding the full model catalog on ordinary inbound replies so Telegram message handling no longer pays multi-second core startup latency before reply generation. Thanks @vincentkoc.
- Gateway/Discord startup: load only configured channel plugins during gateway boot, and lazy-load Discord provider/session runtime setup so startup stops importing unrelated providers and trims cold-start delay. Thanks @vincentkoc.
- Security/exec: harden macOS allowlist resolution against wrapper and `env` spoofing, require fresh approval for inline interpreter eval with `tools.exec.strictInlineEval`, wrap Discord guild message bodies as untrusted external content, and add audit findings for risky exec approval and open-channel combinations.
- Agents/inbound: lazy-load media and link understanding for plain-text turns and cache synced auth stores by auth-file state so ordinary inbound replies avoid unnecessary startup churn. Thanks @vincentkoc.
- Telegram/polling: hard-timeout stuck `getUpdates` requests so wedged network paths fail over sooner instead of waiting for the polling stall watchdog. Thanks @vincentkoc.
- Agents/models: cache `models.json` readiness by config and auth-file state so embedded runner turns stop paying repeated model-catalog startup work before replies. Thanks @vincentkoc.
It flags common footguns (Gateway auth exposure, browser control exposure, elevated allowlists, filesystem permissions).
It flags common footguns (Gateway auth exposure, browser control exposure, elevated allowlists, filesystem permissions, permissive exec approvals, and open-channel tool exposure).
OpenClaw is both a product and an experiment: you’re wiring frontier-model behavior into real messaging surfaces and real tools. **There is no “perfectly secure” setup.** The goal is to be deliberate about:
@ -185,6 +185,7 @@ If more than one person can DM your bot:
- **Inbound access** (DM policies, group policies, allowlists): can strangers trigger the bot?
- **Tool blast radius** (elevated tools + open rooms): could prompt injection turn into shell/file/network actions?
- **Exec approval drift** (`security=full`, `autoAllowSkills`, interpreter allowlists without `strictInlineEval`): are host-exec guardrails still doing what you think they are?
| `gateway.real_ip_fallback_enabled` | warn/critical | Trusting `X-Real-IP` fallback can enable source-IP spoofing via proxy misconfig | `gateway.allowRealIpFallback`, `gateway.trustedProxies` | no |
| `discovery.mdns_full_mode` | warn/critical | mDNS full mode advertises `cliPath`/`sshPort` metadata on local network | `discovery.mdns.mode`, `gateway.bind` | no |
| `config.insecure_or_dangerous_flags` | warn | Any insecure/dangerous debug flags enabled | multiple keys (see finding detail) | no |
| `hooks.token_reuse_gateway_token` | critical | Hook ingress token also unlocks Gateway auth | `hooks.token`, `gateway.auth.token` | no |
| `hooks.token_too_short` | warn | Easier brute force on hook ingress | `hooks.token` | no |
| `hooks.default_session_key_unset` | warn | Hook agent runs fan out into generated per-request sessions | `hooks.defaultSessionKey` | no |
| `hooks.allowed_agent_ids_unrestricted` | warn/critical | Authenticated hook callers may route to any configured agent | `hooks.allowedAgentIds` | no |
| `hooks.request_session_key_enabled` | warn/critical | External caller can choose sessionKey | `hooks.allowRequestSessionKey` | no |
| `hooks.request_session_key_prefixes_missing` | warn/critical | No bound on external session key shapes | `hooks.allowedSessionKeyPrefixes` | no |
| `sandbox.docker_config_mode_off` | warn | Sandbox Docker config present but inactive | `agents.*.sandbox.mode` | no |
| `sandbox.dangerous_network_mode` | critical | Sandbox Docker network uses `host` or `container:*` namespace-join mode | `agents.*.sandbox.docker.network` | no |
| `tools.exec.host_sandbox_no_sandbox_defaults` | warn | `exec host=sandbox` resolves to host exec when sandbox is off | `tools.exec.host`, `agents.defaults.sandbox.mode` | no |
| `tools.exec.host_sandbox_no_sandbox_agents` | warn | Per-agent `exec host=sandbox` resolves to host exec when sandbox is off | `agents.list[].tools.exec.host`, `agents.list[].sandbox.mode` | no |
| `tools.exec.safe_bins_interpreter_unprofiled` | warn | Interpreter/runtime bins in `safeBins` without explicit profiles broaden exec risk | `tools.exec.safeBins`, `tools.exec.safeBinProfiles`, `agents.list[].tools.exec.*` | no |
| `security.exposure.open_groups_with_elevated` | critical | Open groups + elevated tools create high-impact prompt-injection paths | `channels.*.groupPolicy`, `tools.elevated.*` | no |
| `security.exposure.open_groups_with_runtime_or_fs` | critical/warn | Open groups can reach command/file tools without sandbox/workspace guards | `channels.*.groupPolicy`, `tools.profile/deny`, `tools.fs.workspaceOnly`, `agents.*.sandbox.mode` | no |
| `security.trust_model.multi_user_heuristic` | warn | Config looks multi-user while gateway trust model is personal-assistant | split trust boundaries, or shared-user hardening (`sandbox.mode`, tool deny/workspace scoping) | no |
| `tools.profile_minimal_overridden` | warn | Agent overrides bypass global minimal profile | `agents.list[].tools.profile` | no |
| `plugins.tools_reachable_permissive_policy` | warn | Extension tools reachable in permissive contexts | `tools.profile` + tool allow/deny | no |
| `models.small_params` | critical/info | Small models + unsafe tool surfaces raise injection risk | model choice + sandbox/tool policy | no |
| `gateway.real_ip_fallback_enabled` | warn/critical | Trusting `X-Real-IP` fallback can enable source-IP spoofing via proxy misconfig | `gateway.allowRealIpFallback`, `gateway.trustedProxies` | no |
| `discovery.mdns_full_mode` | warn/critical | mDNS full mode advertises `cliPath`/`sshPort` metadata on local network | `discovery.mdns.mode`, `gateway.bind` | no |
| `config.insecure_or_dangerous_flags` | warn | Any insecure/dangerous debug flags enabled | multiple keys (see finding detail) | no |
| `hooks.token_reuse_gateway_token` | critical | Hook ingress token also unlocks Gateway auth | `hooks.token`, `gateway.auth.token` | no |
| `hooks.token_too_short` | warn | Easier brute force on hook ingress | `hooks.token` | no |
| `hooks.default_session_key_unset` | warn | Hook agent runs fan out into generated per-request sessions | `hooks.defaultSessionKey` | no |
| `hooks.allowed_agent_ids_unrestricted` | warn/critical | Authenticated hook callers may route to any configured agent | `hooks.allowedAgentIds` | no |
| `hooks.request_session_key_enabled` | warn/critical | External caller can choose sessionKey | `hooks.allowRequestSessionKey` | no |
| `hooks.request_session_key_prefixes_missing` | warn/critical | No bound on external session key shapes | `hooks.allowedSessionKeyPrefixes` | no |
| `sandbox.docker_config_mode_off` | warn | Sandbox Docker config present but inactive | `agents.*.sandbox.mode` | no |
| `sandbox.dangerous_network_mode` | critical | Sandbox Docker network uses `host` or `container:*` namespace-join mode | `agents.*.sandbox.docker.network` | no |
| `tools.exec.host_sandbox_no_sandbox_defaults` | warn | `exec host=sandbox` resolves to host exec when sandbox is off | `tools.exec.host`, `agents.defaults.sandbox.mode` | no |
| `tools.exec.host_sandbox_no_sandbox_agents` | warn | Per-agent `exec host=sandbox` resolves to host exec when sandbox is off | `agents.list[].tools.exec.host`, `agents.list[].sandbox.mode` | no |
| `tools.exec.security_full_configured` | warn/critical | Host exec is running with `security="full"` | `tools.exec.security`, `agents.list[].tools.exec.security` | no |
| `security.exposure.open_channels_with_exec` | warn/critical | Shared/public rooms can reach exec-enabled agents | `channels.*.dmPolicy`, `channels.*.groupPolicy`, `tools.exec.*`, `agents.list[].tools.exec.*` | no |
| `security.exposure.open_groups_with_elevated` | critical | Open groups + elevated tools create high-impact prompt-injection paths | `channels.*.groupPolicy`, `tools.elevated.*` | no |
| `security.exposure.open_groups_with_runtime_or_fs` | critical/warn | Open groups can reach command/file tools without sandbox/workspace guards | `channels.*.groupPolicy`, `tools.profile/deny`, `tools.fs.workspaceOnly`, `agents.*.sandbox.mode` | no |
| `security.trust_model.multi_user_heuristic` | warn | Config looks multi-user while gateway trust model is personal-assistant | split trust boundaries, or shared-user hardening (`sandbox.mode`, tool deny/workspace scoping) | no |
| `tools.profile_minimal_overridden` | warn | Agent overrides bypass global minimal profile | `agents.list[].tools.profile` | no |
| `plugins.tools_reachable_permissive_policy` | warn | Extension tools reachable in permissive contexts | `tools.profile` + tool allow/deny | no |
| `models.small_params` | critical/info | Small models + unsafe tool surfaces raise injection risk | model choice + sandbox/tool policy | no |
## Control UI over HTTP
@ -528,6 +533,7 @@ Even with strong system prompts, **prompt injection is not solved**. System prom
- Run sensitive tool execution in a sandbox; keep secrets out of the agent’s reachable filesystem.
- Note: sandboxing is opt-in. If sandbox mode is off, exec runs on the gateway host even though tools.exec.host defaults to sandbox, and host exec does not require approvals unless you set host=gateway and configure exec approvals.
- Limit high-risk tools (`exec`, `browser`, `web_fetch`, `web_search`) to trusted agents or explicit allowlists.
- If you allowlist interpreters (`python`, `node`, `ruby`, `perl`, `php`, `lua`, `osascript`), enable `tools.exec.strictInlineEval` so inline eval forms still need explicit approval.
- **Model choice matters:** older/smaller/legacy models are significantly less robust against prompt injection and tool misuse. For tool-enabled agents, use the strongest latest-generation, instruction-hardened model available.
- `tools.exec.security` (default: `deny` for sandbox, `allowlist` for gateway + node when unset)
- `tools.exec.ask` (default: `on-miss`)
- `tools.exec.node` (default: unset)
- `tools.exec.strictInlineEval` (default: false): when true, inline interpreter eval forms such as `python -c`, `node -e`, `ruby -e`, `perl -e`, `php -r`, `lua -e`, and `osascript -e` always require explicit approval and are never persisted by `allow-always`.
- `tools.exec.pathPrepend`: list of directories to prepend to `PATH` for exec runs (gateway + sandbox only).
- `tools.exec.safeBins`: stdin-only safe binaries that can run without explicit allowlist entries. For behavior details, see [Safe bins](/tools/exec-approvals#safe-bins-stdin-only).
- `tools.exec.safeBinTrustedDirs`: additional explicit directories trusted for `safeBins` path checks. `PATH` entries are never auto-trusted. Built-in defaults are `/bin` and `/usr/bin`.
@ -143,6 +144,7 @@ Use the two controls for different jobs:
Do not treat `safeBins` as a generic allowlist, and do not add interpreter/runtime binaries (for example `python3`, `node`, `ruby`, `bash`). If you need those, use explicit allowlist entries and keep approval prompts enabled.
`openclaw security audit` warns when interpreter/runtime `safeBins` entries are missing explicit profiles, and `openclaw doctor --fix` can scaffold missing custom `safeBinProfiles` entries.
If you explicitly allowlist interpreters, enable `tools.exec.strictInlineEval` so inline code-eval forms still require a fresh approval.
For full policy details and examples, see [Exec approvals](/tools/exec-approvals#safe-bins-stdin-only) and [Safe bins versus allowlist](/tools/exec-approvals#safe-bins-versus-allowlist).
"tools.exec.pathPrepend":"Directories to prepend to PATH for exec runs (gateway/sandbox).",
"tools.exec.safeBins":
"Allow stdin-only safe binaries to run without explicit allowlist entries.",
"tools.exec.strictInlineEval":
"Require explicit approval for interpreter inline-eval forms such as `python -c`, `node -e`, `ruby -e`, or `osascript -e`. Prevents silent allowlist reuse and downgrades allow-always to ask-each-time for those forms.",
"tools.exec.safeBinTrustedDirs":
"Additional explicit directories trusted for safe-bin path checks (PATH entries are never auto-trusted).",
title:"Interpreter allowlist entries are missing strictInlineEval hardening",
detail:`Interpreter/runtime allowlist entries were found without strictInlineEval enabled:\n${interpreterAllowlistHits.map((entry)=>`- ${entry}`).join("\n")}`,
remediation:
"Set tools.exec.strictInlineEval=true (or per-agent tools.exec.strictInlineEval=true) when allowlisting interpreters like python, node, ruby, perl, php, lua, or osascript.",