The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to use npx madge to detect circular dependencies, which involves downloading and executing a package from the NPM registry. This is a common and legitimate practice for dependency analysis.
[EXTERNAL_DOWNLOADS]: The use of npx triggers a download of the madge package from registry.npmjs.org, which is a well-known and trusted service for development tools.
[COMMAND_EXECUTION]: The skill utilizes bash commands, specifically grep and npx, to perform static analysis on the source code to identify import patterns and structural issues.
[PROMPT_INJECTION]: The skill handles untrusted codebase data and documentation, which constitutes an indirect prompt injection surface (Category 8).
Ingestion points: The skill reads system codebase files, configurations, and architecture documentation using the Read, Glob, and Grep tools.
Boundary markers: The instructions do not define specific delimiters or security warnings to prevent the agent from following instructions embedded within the codebase being reviewed.
Capability inventory: The agent is granted Bash, WebFetch, and file-reading capabilities, which could be leveraged if malicious instructions in the code were followed.
Sanitization: There is no explicit sanitization or validation of the content retrieved from the codebase files before it is processed by the agent.

review-software-architecture