The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted user data by reading document files from the workspace. This is an inherent surface for indirect prompt injection, where an attacker could place malicious instructions inside a document to influence the agent's output during comparison.
Ingestion points: Reads local .md and .txt files as specified in SKILL.md.
Boundary markers: The skill does not explicitly use delimiters or instructions to ignore commands within the source files, though it focuses on summarizing changes rather than executing content.
Capability inventory: The skill is restricted to file reading and text processing. It lacks network access, shell execution, or file-writing capabilities.
Sanitization: No explicit sanitization or filtering of document content is performed prior to analysis.
[SAFE]: The skill contains no executable scripts, third-party code, or external dependencies. It relies exclusively on natural language instructions for the agent to perform its primary function.

changelog-narrator