The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill relies on external, potentially untrusted data sources to generate technical designs and guide subagent activities.
Ingestion points: The skill reads requirements.md and codebase digests (.skills/<CODE>/scan.md) to determine feature architecture and satisfying requirements.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands within the input files are defined in the skill instructions.
Capability inventory: The agent is authorized to dispatch subagents (scan and review), write files to the project directory (docs/specs/, .skills/), and invoke complex sub-skills like plan-tasks and load-subgraph.
Sanitization: There is no evidence of sanitization or validation of the content extracted from requirements or codebase scans before it is used to formulate design decisions or passed to subagents.
[COMMAND_EXECUTION]: The skill automates the dispatching of specialized "subagents" (scan and review) to perform auxiliary tasks. These secondary agents process the same untrusted input data, potentially leading to coordinated logic manipulation if the inputs contain adversarial content.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for identifying and proposing new third-party dependencies using external tools (e.g., Context7 MCP or research). Although adoption requires user confirmation, the automated research phase is triggered by the analysis of untrusted requirements.

design-solution