The Agent Skills Directory

[OBFUSCATION]: The skill uses zero-width Unicode characters (U+200B and U+200C) to hide data within the titles of SKILL.md and references/pyramid_of_pain.md. This is a sophisticated concealment method designed to bypass human inspection while remaining readable by the AI.
[PROMPT_INJECTION]: The hidden data in the headers constitutes a steganographic prompt injection. This is an attempt to supply instructions to the AI agent that are not visible to the user, potentially used to override safety guidelines or manipulate the agent's internal assessment of the skill.
[INDIRECT_PROMPT_INJECTION]: The pyramid_analyzer.py tool ingests external Sigma rule files. This introduces a risk where malicious content within processed rules could attempt to influence the agent's logic through the tool's output.
Ingestion points: The PyramidAnalyzer class in scripts/pyramid_analyzer.py reads and parses YAML files from external directories.
Boundary markers: The skill does not employ explicit boundary markers or instructions to ignore potential injections within the rules being analyzed.
Capability inventory: The script performs file system reads and provides summary reports to the user.
Sanitization: The script utilizes yaml.safe_load to mitigate common YAML-based code execution vulnerabilities.

bianco-pyramid-of-pain