The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from a wide range of external sources, including CSV files, JSON datasets, and third-party platform data (e.g., Trustpilot, Naver, Coupang). While this creates a potential surface for indirect prompt injection if the source data contains malicious instructions, the skill specifies that data should be quoted as-is, which helps maintain the integrity of the report.\n
Ingestion points: Multiple external files including competitor_matrix.csv, social_mentions.json, and Trustpilot dataset (SKILL.md).\n
Boundary markers: None explicitly defined, though instructions mandate quoting outputs as-is.\n
Capability inventory: Includes local file reading, slide generation via the marp tool, and chart creation using matplotlib or plotly.\n
Sanitization: No specific filtering or sanitization logic is provided for the ingested data.\n- [COMMAND_EXECUTION]: The skill utilizes the marp command-line interface to build PDF and HTML reports from markdown source files. It also uses pdfinfo and grep for verifying the output's page count and checking for design violations. These commands are localized to the workspace and aligned with the skill's primary purpose.\n- [CREDENTIALS_UNSAFE]: The skill includes a clear directive against the exposure of plaintext API keys, which is a positive security practice to prevent accidental credential leakage in generated reports.

marp-slide-build