The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from multiple repository sources and uses it to drive internal logic and generate new file content, creating a vulnerability surface for indirect prompt injection.
Ingestion points: The workflow in references/claude-skill.md reads specifications from design/gdd/*.md, architectural decisions from docs/architecture/*.md, and test plans from production/qa/qa-plan-*.md.
Capability inventory: The skill possesses the ability to create and modify files using the Write tool and to initiate sub-tasks or spawn new agents via the Task tool.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore instructions' warnings when interpolating content from the source files into the agent's working context.
Sanitization: There is no evidence of data sanitization, validation, or escaping logic applied to the external file content before it is processed by the LLM.

create-stories