The Agent Skills Directory

[SAFE]: The skill and its associated files function as a meta-utility for skill engineering. No malicious patterns, such as data exfiltration, credential harvesting, or unauthorized network operations, were found.
[DYNAMIC_EXECUTION]: The provided render-graphs.js utility uses child_process.execSync to invoke the system's Graphviz dot command. This is standard functionality for converting textual graph descriptions into SVG images.
[PROMPT_INJECTION]: The skill includes documentation on using persuasion techniques (e.g., authority, commitment, scarcity) to ensure agents adhere to strict rules (like TDD) and resist bypassing them under pressure. These are pedagogical strategies for instruction resilience and are not intended for bypassing safety filters.
[INDIRECT_PROMPT_INJECTION]: A vulnerability surface exists in render-graphs.js because it ingests and processes content from external skill directories. 1. Ingestion points: Content is read from SKILL.md files by the Graphviz renderer. 2. Boundary markers: The script uses markdown triple-backtick delimiters to identify dot blocks. 3. Capability inventory: The script can execute shell commands (dot) and write SVG files to the local disk. 4. Sanitization: No sanitization is performed on the dot source code before it is passed to the command-line tool.
[COMMAND_EXECUTION]: The documentation includes usage examples for standard command-line tools such as ls, grep, wc, and git for skill file management and discovery.

writing-skills