The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from external web sources and local project files, which could potentially contain malicious instructions designed to influence the design specifications and subsequent implementation plans. * Ingestion points: The skill ingests data from WebFetch, WebSearch, git log, and various project documentation files like CONTEXT.md and ARCHITECTURE.md. * Boundary markers: The skill uses structured templates for agent prompts and report formats, which provides some separation, but no specific delimiters are used to wrap the ingested content. * Capability inventory: The skill can write files and initiate implementation workflows by invoking other platform skills (e.g., /plan --auto). * Sanitization: There is no evidence of explicit sanitization or escaping of the ingested data before it is interpolated into agent prompts.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute git log --oneline -30. This is a restricted and legitimate use of shell commands for gathering project context.
[EXTERNAL_DOWNLOADS]: The skill fetches framework documentation and industry best practices using WebSearch, WebFetch, and the context7 MCP. These operations are part of the intended research capability and target well-known documentation sources.

brainstorm