The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection because its core functionality involves reading and processing untrusted data from project files to inform its output.\n
Ingestion points: The skill explicitly instructs the agent to read various project files, including package.json, pyproject.toml, Cargo.toml, Dockerfile, and existing README.md files (as described in the 'Intake' section of SKILL.md).\n
Boundary markers: There are no instructions for the agent to use protective delimiters or to disregard instructions found within the processed project files.\n
Capability inventory: The skill utilizes Write and Edit tools (defined in the SKILL.md frontmatter), which allow it to modify the filesystem based on the output of its analysis.\n
Sanitization: The skill lacks any steps for sanitizing, filtering, or escaping content read from external project files before processing it with the language model.

readme-writer