The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file employs the dynamic command execution syntax to run cat and find. These commands are used to gather context about the project's configuration and existing transform files. This is a legitimate use case for a developer tool and does not present a security risk.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for building transformations that process external data from GraphQL queries. This establishes an attack surface for indirect prompt injection if the ingested data contains malicious instructions.\n
Ingestion points: Data is fetched via GraphQL queries and passed to Python or Jinja2 transformation logic as described in rules/python-transform.md and rules/jinja2-template.md.\n
Boundary markers: The instructions do not specifically require the use of boundary markers or delimiters for untrusted data values.\n
Capability inventory: The skill uses tools including Bash, Write, Edit, Read, and Grep to manage transformation files and configurations.\n
Sanitization: While not universally mandated, the skill includes specific examples of data sanitization, such as XML entity escaping in examples.md for SVG generation.

infrahub-managing-transforms