The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from external design documents and source code implementation files to verify implementation claims and apply synchronization updates.
Ingestion points: The skill reads local design.config.json, design markdown documents (e.g., in .claude/design/), and codebase source files.
Capability inventory: The skill is configured to use Bash and Edit tools to apply synchronization changes to the codebase and update documentation frontmatter (e.g., last-synced timestamps).
Risks: Malicious instructions embedded within design documents or source code comments (e.g., hidden implementation claims) could potentially influence the agent's behavior during the analysis phase or during the application of updates, leading to unintended file modifications or command execution.
Mitigation: The skill workflow includes a manual review phase for 'unsafe' updates and generates a sync report before applying changes, providing a human-in-the-loop checkpoint.

design-sync