The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process arbitrary prompts provided by users to extract reusable elements.\n- Ingestion points: User-supplied prompts and design system specifications are processed as described in SKILL.md and implemented in modules/library_updater.md.\n- Boundary markers: The skill lacks explicit delimiters or instructions for the agent to ignore potentially malicious instructions embedded within the source prompts being analyzed.\n- Capability inventory: Across modules/library_updater.md and SKILL.md, the skill is capable of writing to a local SQLite database (extracted_results/elements.db), JSON files, and markdown reports. It does not possess network access or shell execution capabilities.\n- Sanitization: While the element_extractor.md module performs quality checks for clarity and length, it does not include specific sanitization filters to detect or neutralize adversarial prompt injection payloads within the processed data.

universal-learner