The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests data from external URLs to extract brand DNA (colors, typography, and voice). This content is interpolated into prompts to generate slide decks, which presents a risk of indirect prompt injection if the source website contains instructions designed to override the agent's behavior.
Ingestion points: Data extracted via Firecrawl from arbitrary URLs is processed in brand_extractor.py and referenced in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard embedded instructions within the scraped brand data.
Capability inventory: The skill creates and writes local HTML files (slides.html) and provides patterns for executing CLI commands.
Sanitization: No sanitization logic for the scraped content is present; parameters are passed directly to the generation prompt.
[COMMAND_EXECUTION]: Documentation for the skill includes a Python script (deck_generator.py) that uses subprocess.run to execute the claude CLI tool with variable-based prompts. This pattern is noted as a surface for command injection if implemented in environments with unvalidated user input.
[EXTERNAL_DOWNLOADS]: The installation instructions direct users to clone the skill's source code from an external GitHub repository (ItsssssJack/power-design).

power-design-slides