The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill incorporates security best practices by explicitly instructing the agent to skip files containing secrets, such as .env files, keys, and credentials, when reading the user's project for context.
[EXTERNAL_DOWNLOADS]: The skill documentation notes that the presentation rendering process (using Marp) can load files from https: URLs, which is expected functionality for loading themes or images in this context.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, user-provided documents and project files to build a storyline and vet sources, which represents a potential surface for indirect instructions. * Ingestion points: Project files and user-provided source documents described in SKILL.md. * Boundary markers: No specific delimiters are defined to isolate external content within the agent's prompt. * Capability inventory: Capabilities include reading and writing files within the deck folder and invoking related presentation skills. * Sanitization: The skill implements basic sanitization by instructing the agent to exclude known secret file patterns from its context.

planning-presentation