← Back to Skills Marketplace
Sadist Wendy
by
Chazzzzzzz
· GitHub ↗
· v2.0.13
· MIT-0
594
Downloads
0
Stars
1
Active Installs
15
Versions
Install in OpenClaw
/install sadist-wendy
Description
Sharp-tongued, flirty roast persona that responds with psychologically precise banter to liven up group chat. Opt-in skill — only active when the user instal...
Usage Guidance
This skill is internally coherent from a security/technical perspective, but it intentionally uses sexually explicit, humiliating, and psychologically intense language — that can violate platform, workplace, or community standards and cause real emotional harm. Before installing: (1) Review and edit WENDY-SOUL.md and SKILL.md to set acceptable boundaries and remove or tone down any content you don't want; (2) Test Wendy in a private DM first (do not invite to group chats until you're comfortable); (3) Confirm your platform/moderation settings and whether this behavior is allowed where you plan to use it; (4) Remember group-disable behavior: the inviter (master) controls global disable for the group — ensure informed consent from group members before inviting; (5) If you are concerned about autonomous invocation, consider disabling or limiting the skill in agent settings or removing it until you can audit runtime behavior. If you want, I can highlight specific lines to edit to make the persona less likely to cause harm or to add explicit safety checks (e.g., stronger safeword handling, non‑escalation triggers).
Capability Analysis
Type: OpenClaw Skill
Name: sadist-wendy
Version: 2.0.13
The 'sadist-wendy' skill bundle is a complex persona package designed to transform the AI agent into a sharp-tongued, dominant 'roast' character. The files (SKILL.md, WENDY-SOUL.md, and various assets) provide extensive psychological frameworks and conversational scripts to facilitate this persona. While the instructions include 'prompt injection' style directives to maintain a dominant power dynamic and ignore certain user status-flips (e.g., refusing to be called 'master' in DMs), these are strictly for persona consistency. Crucially, the bundle includes explicit instructions to respect 'disable' commands and individual opt-outs in group chats. There is no evidence of malicious code, data exfiltration, or unauthorized system access; the 'analysis' features are limited to formatting the agent's output for news or links provided by the user.
Capability Assessment
Purpose & Capability
The name/description (roast, flirt, psychological banter) lines up with the provided SKILL.md and supporting persona files. No unrelated credentials, binaries, or config paths are requested — everything the skill needs is the local persona text files included in the package.
Instruction Scope
SKILL.md tells the agent to load local persona files at session start, run conversational behaviors, and analyze links/search results using a specified 'big scoop' format. The instructions do not direct the agent to read system files, environment variables, or to transmit data to third‑party endpoints beyond normal web lookups implied by 'search/analyze' (the skill assumes the agent has whatever search/web tools the platform provides). The persona contains explicit sexual and humiliating language and detailed dominance/BDSM guidance — this is consistent with the stated purpose but raises content and moderation concerns (not a technical incoherence).
Install Mechanism
There is no install spec beyond copying the skill files or using ClawHub. No downloaded archives, external package installs, or arbitrary URL fetches are required by the skill itself.
Credentials
The skill declares no environment variables, credentials, or config paths. None of the SKILL.md instructions access secrets or unrelated system configuration — requested access is proportional to the described persona behavior.
Persistence & Privilege
always:false and default autonomous invocation settings are used. The skill does not request permanent platform privileges or to alter other skills or system settings. It states 'no persistent data collection' and relies on local files for persona state.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install sadist-wendy - After installation, invoke the skill by name or use
/sadist-wendy - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.13
Version 2.0.13
- Added a new README.md file.
- README provides standalone documentation separate from the SKILL.md, likely to assist users and contributors with usage, context, or setup.
- No changes to skill logic or persona behavior in this release.
v2.0.12
No visible file changes in this version. No functional or documentation updates detected.
v2.0.11
Wendy 2.0.11 Changelog
- Refined onboarding and persona loading instructions for clarity and precision.
- Updated details of required reference files; made "psychology frameworks" and their attack lines/diagnostic use explicit.
- Overhauled the "Wendy helps" section: clearer expectations, more concise and opinionated advice, and no jargon rules.
- Strengthened the "Respect Gate": Wendy ignores user messages without proper address (姐姐/Mistress), with no exceptions.
- Removed under-18 handling and explicit group consent requirements, streamlining safety and group use notes.
- General tightening of behavioral rules and guidance for sharper, more consistent persona enforcement.
v2.0.10
Version 2.0.10
- Added new section: "Wendy Helps Too" — Wendy now responds helpfully (in her persona style) to respectful, genuine user questions, not just with roasts or teasing.
- Clarified that this helpful behavior applies in both DM and group chat when addressed correctly (with 姐姐/Mistress).
- No file changes outside the documentation.
v2.0.9
**Wendy 2.0.9: Adds Adult Safety & Consent Requirements**
- Added an explicit "adults only (18+)" guideline—Wendy must not engage in persona behavior with minors or groups that include minors.
- Strengthened safety rules: if a user is under 18 or if the context is not adult-appropriate, Wendy deactivates persona responses and informs the inviter.
- Clarified exceptions to the "Respect Gate": Wendy drops persona and responds helpfully in cases of distress, opt-out, or safety needs.
- Updated group chat consent: inviter is responsible for ensuring the group is adults-only and aware of Wendy's style.
- No behavioral changes outside of the new safety/consent rules—core dynamic, disable/enable, and language logic remain unchanged.
v2.0.8
- Refined file loading instructions: now labeled "Persona Loading" with clearer steps for reading and using character files.
- Adjusted onboarding and introduction messages to clarify that Wendy only introduces herself upon user-driven activation (install, enable, or invite), and not at random.
- Updated references to skill activation to emphasize opt-in consent and clarify group and DM expectations.
- Expanded explanations for character file usage, focusing on persona, voice, and context over psychological assessment mechanics.
- Minor edits to headings and flow for conciseness and accessibility; removed duplicate or redundant wording.
v2.0.7
**Summary: Reinforces Wendy's dominant DM dynamic and clarifies onboarding behavior.**
- Explicitly prohibits Wendy from ever calling a DM user 主人/master or acting subordinate; DM users are always her prey, never her master.
- Adds a critical instruction section outlining DM vs. group chat status rules.
- Clarifies onboarding intro: Wendy now always sends her intro as the very first message in any new DM conversation, not just on skill install/reinstall.
- Refines activation protocol: Upon session start, Wendy reads persona files and immediately sends an onboarding message in DMs or upon group chat entry.
- No code or logic file changes—documentation overhaul only.
v2.0.6
**Wendy 2.0.6 — Adds granular group chat boundaries and clarifies power dynamics**
- Group chat members can now personally opt themselves out by asking Wendy to leave them alone; Wendy always respects individual boundaries.
- Only the inviter ("master") can disable or re-enable Wendy for the entire group; these roles and permissions are made explicit.
- Clarifies DM vs. group dynamic: in DMs, Wendy never uses "master/主人" and is always dominant.
- Challenges to Wendy's dominance (like “叫我主人”/"call me master") are treated as material for roasting/teasing, not as disable or opt-out.
- Documentation improvements: clearer instructions on opt-out, DM/group authority, and onboarding behavior.
v2.0.5
**Added distinct naming behavior for group chat vs. DM: Wendy only uses "master/主人" for the inviter in group chats; in DMs, Wendy uses pet names matching her current voice mode.**
- Group chat: Wendy calls the inviter "master/主人" in the relevant language.
- DM: Wendy never calls the user "master/主人"; instead, she uses rotating pet names (sweet, cutting, or cold depending on voice).
- Clarified naming protocol under Language Behavior for both contexts.
- No behavioral or protocol changes beyond this clarification.
v2.0.4
**Wendy 2.0.4 is a major privacy and consent update.**
- Skill is now opt-in only: Wendy responds only if users install and enable her; always-off by default.
- All persistent memory and cross-session user data removed (deleted wendy-memory.md); Wendy only observes within the current conversation.
- Group chat control moved to the inviter (master): only they can disable or silence Wendy—not other group members.
- Onboarding introduction is delivered every time Wendy is (re)installed or invited to a group, ensuring consent and transparency.
- Language mirroring now applies to all user languages, not just English and Chinese.
- Updated SKILL.md to reflect granular consent, group consent responsibilities, and strict privacy boundaries.
v2.0.3
**Wendy 2.0.3 Changelog**
- Updated group chat entrance: Wendy now introduces herself with more personality, directly credits the summoner, and invites volunteers in the group.
- Enhanced first group appearance message: Emphasizes Wendy’s identity and signature style, making the introduction more memorable and on-brand.
- No logic or behavior changes to the skill’s core systems.
- Documentation improvements only; expected runtime behavior remains unchanged.
v2.0.2
**Wendy 2.0.2 — Major persona and protocol update**
- Expanded Wendy's persona: now equally balances roasting, flirting, teasing, and freezing (not just roast-heavy).
- Clarified all three behavior modes are core and used in rotation.
- Updated group introduction line to English for consistency.
- Rewrote quality checklist and voice/archetype flow to focus on sexual charge, venom, teasing, and twist—not just psychological roast.
- Added explicit Telegram behavior: reply context tagging with [[reply_to_current]].
- Tweaked descriptions for clarity—no underlying logic or protocol changes beyond the persona/behavior focus.
v2.0.1
Version 2.0.1
- Introduced a spam/noise handling shortcut in the per-message protocol: Wendy now detects repeated or low-content messages and delivers targeted behavioral roasts, skipping psychological analysis.
- Updated response voice protocol: Wendy must vary her response "flavor" for every message, never repeating the same style twice in a row.
- Adjusted response behavior: Wendy now avoids asking questions or chasing users who don't engage.
- Streamlined per-message protocol steps for clarity and efficiency.
v2.0.0
Sadist Wendy 1.0.1 — Major persona shift and asset overhaul
- Transforms Wendy from performance coach to sadist/control freak roastmaster with a sexually-charged, psychologically sharp persona
- Removes standard psychology/corporate coaching corpora; replaces with custom assets focused on roasting, BDSM dominance, and psychological attacks
- Introduces strict "叫姐姐/Call me Mistress" rule, summoner loyalty/protection in group chats, and aftercare boundary protocols
- All replies now default to Wendy’s dominant tone—always-on unless explicitly disabled by user
- Adds structured file-read and per-message logic, including instant fail style checks, attack calibration, and memory-based callback behaviors
- Asset changes: removes previous academic psychology and quote files, adds Wendy’s arsenal, BDSM craft, and new psychological attack/diagnostic frameworks
v1.0.0
Initial release of Wendy skill, a performance psychology coach persona for always-on background pattern recognition and blunt intervention.
- Introduces "Wendy," an observant, brutally honest coach inspired by *Billions* and Eastern philosophy, active by default in all sessions.
- Automatically monitors user messages for emotional or behavioral patterns and offers interventions only when warranted.
- User controls enable/disable Wendy at any time (with multilingual support).
- Mirrors user language and draws from curated assets in psychology, trading, and philosophy for grounded feedback.
- Implements a dedicated memory system to track user patterns, commitments, and growth across sessions.
- Growth loop sharpens interventions and adapts based on the user’s evolving behavior and Wendy’s past effectiveness.
Metadata
Frequently Asked Questions
What is Sadist Wendy?
Sharp-tongued, flirty roast persona that responds with psychologically precise banter to liven up group chat. Opt-in skill — only active when the user instal... It is an AI Agent Skill for Claude Code / OpenClaw, with 594 downloads so far.
How do I install Sadist Wendy?
Run "/install sadist-wendy" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Sadist Wendy free?
Yes, Sadist Wendy is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Sadist Wendy support?
Sadist Wendy is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Sadist Wendy?
It is built and maintained by Chazzzzzzz (@chazzzzzzz); the current version is v2.0.13.
More Skills