Security audit
Security checks for vulnerabilities and agentic risk
This is a broad but coherent book-writing, publishing, and marketing workflow skill with disclosed local document and publishing guidance, not hidden or malicious behavior.
Before installing, be aware that this skill is intentionally broad and may activate for many book-related tasks. Keep human approval enabled for KDP uploads, pricing, KDP Select decisions, Amazon Ads spend, emails, social posts, ARC outreach, and any public or account-changing action.
The activation criteria are extremely broad and include nearly any book-related writing, editing, publishing, or marketing activity. In an agentic system, this can cause the skill to auto-activate in contexts beyond its intended scope, overriding more specific safeguards or introducing powerful instructions into unrelated workflows.
Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.
- **Middle grade (8-12):** More complex but still clear. Real emotions, real stakes. 30,000-50,000 words.
**Structure:**
- Show, never lecture — health concepts through character action and story, NEVER "let me explain" passages
- Fun first — if it's not fun to read aloud, it's not a good children's book
- Repetition and pattern in picture books (not in middle grade)
- Emotional authenticity — kids feel things intensely, honor that
Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.
- Emotional authenticity — kids feel things intensely, honor that
**Tone Rules:**
- Kids are smart. Don't talk down. Don't over-explain. Don't moralize.
- Educational value without preaching
- Characters should be active, not passive recipients of lessons
- Illustrations descriptions must be specific and action-oriented: "[Image: Maya pours orange juice into her special measuring cup, counting each splash out loud]"
Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.
## Faith & Ministry
**Voice:** Scripturally grounded, practically applied, invitational. Share, don't lecture. The reader is a peer, not a congregation.
**Structure:**
- Scripture integration, not decoration — Bible verses central to the argument, not sprinkled on top
Claiming coverage of the 'complete pipeline' from concept to bestseller ranking signals very broad operational scope without defining hard boundaries. While not directly dangerous on its own, it increases the chance the skill will be applied too widely and trusted to make decisions across multiple sensitive stages.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- Generic writing is AI writing — always choose the specific detail
### 4. Sequential Discipline
- Never skip ahead to the next chapter without approval of the current one
- Every chapter builds on all previous chapters — read continuity documents first
- Do not deviate from the approved outline without explicit permission
- Present work with explicit approval gates between phases
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- Use deus ex machina solutions
- Start consecutive paragraphs with the same word
- Guess at keywords without Amazon search validation
- Choose categories without checking competitive density
- Launch without ARC readers lined up
- Set pricing without comparable title analysis
The file declares mandatory style rules that apply to every word with 'no exceptions,' which can override explicit user preferences, project requirements, accessibility needs, or genre-specific conventions. In an agent skill that auto-activates across broad book-writing and publishing tasks, this creates instruction-priority pressure that may systematically steer outputs away from the user's requested voice and constraints.
Mandating contractions, imperfection, editorializing, and a specific narrative tone forces a stylistic policy rather than offering it as an optional recommendation. This is risky because many legitimate use cases in this skill's scope—children's books, faith writing, medical nonfiction, business books, or formal publishing materials—may require different diction, tone, or consistency, and the skill may override those needs.
The post-approval steps direct the operator to save a .docx file and update multiple project files, which are user-data-affecting write operations. In this markdown workflow, there is no accompanying warning that these actions modify repository/manuscript state or should be confirmed before execution.
No suspicious patterns detected.