The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates a workflow that generates an 'implementation-prompt.md' file, which is intended to be consumed by other AI agents to execute project builds. This creates an indirect prompt injection surface where instructions or patterns found in ingested legacy code or project documentation could influence the behavior of the downstream agent. Ingestion points: 'references/legacy-extraction.md' (reads existing codebase and manifests) and 'SKILL.md' (captures product decisions). Boundary markers: The skill does not explicitly mandate delimiters or 'ignore' warnings for the generated prompt content. Capability inventory: The skill uses 'rg' for file searching and requires read access to project files. Sanitization: No explicit sanitization of ingested content before prompt interpolation is defined.- [COMMAND_EXECUTION]: The skill references the use of 'rg' (ripgrep) and 'wc' (word count) within 'references/design-doc-rules.md' for consistency checks, such as identifying stale configuration or counting lines. These tools are used for static analysis purposes and do not involve executing arbitrary code from untrusted sources.- [DATA_EXPOSURE]: The skill includes positive security constraints in 'references/legacy-extraction.md', specifically instructing the agent not to copy secrets, credentials, or private endpoints from legacy projects into new designs.

project-design-bootstrap