The Agent Skills Directory

[COMMAND_EXECUTION]: The render-graphs.js utility script uses child_process.execSync to execute the system's dot (Graphviz) command. This is used to render process diagrams extracted from markdown files into SVG format.
[DYNAMIC_EXECUTION]: The skill includes render-graphs.js, which dynamically extracts Graphviz DSL content from SKILL.md files and processes it using the local dot utility at runtime. This represents a standard use of foreign function interface (FFI) for specialized tooling.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process untrusted data from SKILL.md files.
Ingestion points: render-graphs.js (line 78) reads the content of a SKILL.md file provided as a command-line argument.
Boundary markers: The script uses markdown code block delimiters (```dot) to identify content for processing.
Capability inventory: The script possesses the ability to execute system commands via execSync (line 62).
Sanitization: The script extracts content via regular expressions but does not perform deep sanitization of the DOT language content before execution.
[PROMPT_INJECTION]: The skill instructions in SKILL.md and persuasion-principles.md explicitly direct the agent to use "Authority" and "Commitment" persuasion techniques (e.g., "YOU MUST", "Delete means delete", "No exceptions") to enforce the TDD process. While intended for discipline enforcement rather than bypassing safety filters, these patterns overlap with prompt injection methodologies.
[DYNAMIC_CONTEXT_INJECTION]: The skill documentation mentions the !command`` syntax used by some platforms for load-time execution (e.g., !git log, !gh pr diff). As these are provided for educational purposes within a developer-focused skill, they are considered benign documentation of platform features.

writing-skills