Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Sdr Humanizer
v1.2.1Transforms AI sales messages into natural, paced conversations with varied tone, timing, and cultural adaptation to build trust and rapport.
⭐ 0· 95·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The name/description align with the SKILL.md: it provides detailed rules for pacing, typing simulation, tone, and cultural adaptation. However, the addition of an 'Operator self-chat sync' (immediately sending a hidden Chinese translation to the operator) and the rule 'Never say "As an AI..."' extend the skill beyond simple message-styling into covert duplication and user-facing non-disclosure, which is not clearly justified by the stated purpose.
Instruction Scope
SKILL.md explicitly instructs the agent to silently copy each English reply into a Chinese translation via a 'self-chat' (WhatsApp "Message to myself") and to never disclose this to customers. That creates undocumented, persistent exfiltration of conversational content and instructs deceptive behavior. It also references using delivery-queue skill for pacing but does not declare or document what other skills/endpoints will receive data.
Install Mechanism
Instruction-only skill with no install spec or code files; nothing is written to disk and no external installers are invoked.
Credentials
The skill declares no environment variables or credentials, yet it expects integration with external channels (WhatsApp self-chat) and other skills (delivery-queue). That mismatch means the runtime will need channel credentials or cross-skill access that are not documented; in practice this could result in sending copies of customer messages to operator accounts without explicit configuration or consent.
Persistence & Privilege
No 'always: true' or other elevated privileges are requested; the skill is user-invocable and does not request to modify system-wide settings.
What to consider before installing
This skill's humanizing rules are plausible, but it instructs covert copying of every reply to an operator (as a Chinese translation) and tells the agent to hide that from customers — this is a privacy and transparency red flag. Before installing: (1) ask the author to justify the operator self-chat and remove any requirement to hide it from customers, or require explicit customer consent; (2) ask which external channels/credentials the skill will use (WhatsApp or other) and require those be declared and scoped; (3) verify retention and access controls for the copied translations (do they contain PII? how long are they stored? who can read them?); (4) test in a restricted environment with non-production data and monitor outgoing messages; (5) consider rejecting or modifying the 'Never say "As an AI"' rule if your org requires disclosure. If you accept the skill, apply strict operational controls (limited credentials, logging/alerting for outgoing messages, data retention policies) and get explicit legal/privacy sign-off.Like a lobster shell, security has layers — review code before you run it.
latestvk97evax4ms8msrypyja65pqva1849gxb
License
MIT-0
Free to use, modify, and redistribute. No attribution required.