The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from project configuration files and source code during its auditing and detection procedures, which could be exploited to provide malicious instructions to the agent.
Ingestion points: The skill reads external files such as package.json, Cargo.toml, pyproject.toml, CONTRIBUTING.md, and various source files during repo type detection and the path portability audit as described in SKILL.md and references/path-portability.md.
Boundary markers: The instructions do not mandate the use of explicit delimiters or "ignore embedded instructions" warnings when processing the content of these external files.
Capability inventory: According to the allowed-tools in SKILL.md, the skill has access to Write, Edit, and restricted Bash commands including grep, git, find, and ls for auditing and scaffolding tasks.
Sanitization: There is no explicit requirement for the agent to sanitize, validate, or escape the content read from these files before using it to make structural decisions or performing file operations.

project-conventions