The Agent Skills Directory

[SAFE]: The skill implements a standard software development lifecycle (SDLC) for agent skills, focusing on scoping, research, evaluation-driven development, and validation.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it processes external research data and local files to generate new code. However, this behavior is essential to the skill's purpose as an authoring tool and includes a human-review requirement before committing changes.
Ingestion points: Reads local content from the .local/ directory and performs web searches to gather documentation and examples.
Boundary markers: No explicit delimiter instructions are provided for separating external data from the prompt context during the research phase.
Capability inventory: The skill is authorized to write to the local file system (creating skills, updating READMEs) and execute the make check command for validation.
Sanitization: No specific content filtering or sanitization steps are defined for the data ingested during the research workflow.
[COMMAND_EXECUTION]: The skill uses make check for validating the generated skill's integrity. This is a standard development practice and does not involve arbitrary command injection or privilege escalation.

add-skill