The Agent Skills Directory

[SAFE]: No security threats were identified in the skill's instructions or metadata. The workflow emphasizes ethical storytelling and accuracy, with no evidence of credential theft, remote code execution from unknown sources, or persistence mechanisms.
[INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting external data, which is an inherent surface for indirect prompt injection.
Ingestion points: Phase 1 (Research & Validation) involves processing external source material, including books, articles, and public discourse (SKILL.md).
Boundary markers: The workflow uses a 'Hard Precondition' (Phase 0: Reader Briefing) and manual critical review steps (Phase 3) to anchor the agent's output to specific goals rather than raw input instructions.
Capability inventory: The skill's capabilities are restricted to text generation, hyperlink management, and metadata preparation.
Sanitization: Includes a dedicated anonymization protocol (Ethical Storytelling section) to identify and remove fingerprinting details from processed examples.
[DYNAMIC_EXECUTION]: The workflow mentions the use of a local Python script for automated batch quality checks.
Evidence: Phase 4.4 references scripts/corpus_repetition.py (a tool within the sibling synthesis-content-quality skill) to compute repetition and token concentration across article batches.

synthesis-article-writing