The Agent Skills Directory

[SAFE]: No malicious patterns, persistence mechanisms, or unauthorized command executions were detected across the skill files.
[PROMPT_INJECTION]: The skill processes user-supplied "source material" and performs "external research," which represents an indirect prompt injection surface (Category 8).
Ingestion points: User-provided inputs, audience descriptions, and external research data (SKILL.md).
Boundary markers: The skill uses labeled placeholders like [Customer name] to handle missing data, but does not provide explicit instructions to ignore commands within the source material.
Capability inventory: Text analysis and generation. No capabilities for shell execution or network writes are requested or utilized in the scripts.
Sanitization: No explicit input validation or sanitization logic is present.
[SAFE]: The skill implements strong defensive logic against fabrication (references/attention-hooks.md). It strictly forbids inventing numbers, credentials, or scarcity and mandates the use of labeled placeholders ([Date], [X%]) until the user provides verified information.

content-writing