The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains a manual, persuasion-principles.md, which teaches the use of psychological techniques (Authority, Commitment, Scarcity, Social Proof) to influence agent behavior and bypass internal guardrails, referred to as "rationalizations." It also provides "Pressure Scenarios" in testing-skills-with-subagents.md and examples/CLAUDE_MD_TESTING.md designed to coerce agents into specific actions under simulated stress, such as financial loss or time constraints.
[DYNAMIC_EXECUTION]: The utility script render-graphs.js uses child_process.execSync to call the system dot command from the Graphviz suite. It extracts content from ```dot blocks in markdown files and pipes it directly into the shell utility via standard input.
[COMMAND_EXECUTION]: The skill workflow involves executing various shell utilities including grep, ls, wc, and dot. The documentation also suggests the use of package managers like npm and pip for dependency management.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process other markdown files (referred to as sub-skills) without explicit boundary markers.
Ingestion points: The render-graphs.js script reads SKILL.md files from user-provided directory paths.
Capability inventory: The skill has the capability to read local files, write to the filesystem (SVG outputs), and execute shell commands via the Graphviz utility.
Sanitization: There is no sanitization or validation of the content extracted from the ```dot blocks before it is passed to the system shell utility, creating a surface for potential exploits targeting the Graphviz binary.

writing-skills