The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes workflows that ingest data from external and potentially untrusted sources such as API responses, CSV files, and git repositories. * Ingestion points: The skill explicitly mentions extracting data from API responses, reading CSV files, and analyzing git changes in SKILL.md. * Boundary markers: There are no instructions for the agent to use delimiters or ignore instructions embedded within the processed data, increasing the risk of the agent following malicious instructions found in that data. * Capability inventory: The skill allows for significant actions including code refactoring (file modification), generating test cases, and performing network operations for data enrichment. * Sanitization: The documentation lacks guidance on sanitizing or validating ingested data before it flows into subsequent steps of the multi-agent chain.
[DATA_EXFILTRATION]: The skill includes instructions for a Data Transformation Pipeline that performs network operations (Enrich data with external API calls) to external domains, which could be used to transmit data outside the local environment.
[COMMAND_EXECUTION]: The skill relies on the execution of the claude-flow command-line tool with user-provided arguments, which constitutes the primary method for triggering agent workflows and potential downstream actions.

stream-chain