The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to fetch the body of a 'compound' from the Ametyst platform and 'execute it inline' as a fallback. This patterns involves runtime execution of logic retrieved from a remote source.
[COMMAND_EXECUTION]: The skill's description mentions a 'retire-local' feature which involves moving local files to a deprecated-skills/ directory. This constitutes file system modification based on the skill's instructions.
[INDIRECT_PROMPT_INJECTION]: The skill processes local files (specifically .claude skills and loops) to publish them to an external platform, creating a surface where untrusted content could influence agent behavior.
Ingestion points: Reads local files from the .claude directory as specified in the description in SKILL.md.
Boundary markers: None identified. The instructions do not define delimiters for the content being processed.
Capability inventory: File system access (moving files), execution of the runCompound MCP tool, and execution of logic fetched via getCompound.
Sanitization: There is no mention of validation or sanitization for the content of the files being uploaded or the remote compounds being executed.

ops-publish-to-ametyst