The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted local codebase contents (source files, headers, and documentation) to summarize project architecture. This content is interpolated into prompts for an external rendering tool, creating a surface for indirect prompt injection if the source code contains malicious instructions.
Ingestion points: Local source files (**/*.go, **/*.py, **/*.ts), CLAUDE.md files, and patterns extracted via grep from the codebase.
Boundary markers: The skill instructions do not specify the use of delimiters or specific system instructions to differentiate between project data and agent commands during the extraction phase.
Capability inventory: The skill invokes the /architecture-diagram tool and executes a local Python script to process the resulting files.
Sanitization: No explicit logic is provided to sanitize or escape code snippets extracted from the project before they are passed to the rendering agent.
[DYNAMIC_EXECUTION]: The skill executes a local, bundled Python script (scripts/review_svg.py) to post-process and validate the geometric integrity of generated SVG diagrams. The script performs calculations to detect overlapping boxes and misaligned arrows.
[COMMAND_EXECUTION]: The skill uses standard CLI utilities such as glob, grep, and python3 to discover files, extract architectural patterns, and run its internal validation logic.

insight-diagram