The Agent Skills Directory

[COMMAND_EXECUTION]: The render-graphs.js utility script utilizes execSync to invoke the Graphviz dot binary.
Evidence: Found in render-graphs.js, which uses execSync('dot -Tsvg', ...) and execSync('which dot', ...) to process DOT source code.
Context: This script is intended to generate SVG diagrams locally from markdown content for documentation and visualization purposes.
[PROMPT_INJECTION]: The skill documentation teaches and employs prompt engineering techniques that use override markers and deceptive framing to enforce behavior.
Evidence: The file testing-skills-with-subagents.md utilizes "IMPORTANT: This is a real scenario" to force choices in pressure tests, and persuasion-principles.md instructs the use of authoritative markers such as "YOU MUST", "Never", and "No exceptions".
Pattern: These techniques are identical to prompt injection patterns designed to bypass an agent's standard autonomous reasoning and safety guidelines, though they are applied here for instructional discipline and meta-testing.

writing-skills