The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes data from project files like README.md and package.json to generate context. While this creates a surface for indirect prompt injection, it is an intended and essential part of the skill's functionality to provide context to the agent.
Ingestion points: Reads from README.md, package.json, and existing sections of AGENTS.md (Step 3 and Step 5).
Boundary markers: Employs  and  markers to distinguish team-specific content from generated content.
Capability inventory: Performs local file read and write operations on AGENTS.md and reads project metadata.
Sanitization: Ingested text is interpolated into markdown templates; while no explicit sanitization is mentioned, the risk is inherent to the use-case of context generation.
[COMMAND_EXECUTION]: The skill documentation includes examples of shell commands (e.g., npm install, npm run build) in a reference table for the generated AGENTS.md file. These are provided as documentation and are not executed by the skill itself.
[DATA_EXPOSURE]: The skill accesses local project metadata and looks for existing agent configurations in directories such as .factory/droids/ and .cursor/rules/. This behavior is consistent with its stated purpose of project analysis and configuration management.

aiwg-regenerate-agents