The Agent Skills Directory

[COMMAND_EXECUTION]: The file scripts/render-dot.py contains a relative path string (../../../scripts/render-dot.py). The skill's main instructions in SKILL.md direct the agent to execute this file with arguments, which represents a path traversal pattern to access and run code outside the skill's directory.
[PROMPT_INJECTION]: The skill facilitates the creation and optimization of other AI instructions, creating an indirect prompt injection surface where untrusted data could be persisted into skill definitions.
Ingestion points: Phase 1 and Phase 2 process user-supplied scenario examples and existing skill files.
Boundary markers: The skill does not provide explicit delimiters or warnings to ignore instructions embedded in the data it processes.
Capability inventory: The skill involves file writing and command execution through the render-dot.py script.
Sanitization: There is no mechanism described for sanitizing or escaping content before incorporating it into new skill files.
[EXTERNAL_DOWNLOADS]: The ATTRIBUTION.md file links to external resources including documentation from Anthropic and OpenAI, as well as developer blogs, for informational purposes.

optimize-skills