The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's metadata contains an installation command that fetches the SKILL.md file from a GitHub repository using curl.\n- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands for deployment purposes, specifically make install-agents and lib/bin/install-agents, to synchronize configurations to provider-specific directories.\n- [PROMPT_INJECTION]: The skill's auditing and validation workflows ingest markdown files (agents/*.md) from the workspace. This creates a vulnerability surface for indirect prompt injection, as malicious instructions within these audited files could influence the agent's behavior during processing.\n
Ingestion points: Processes agent definition markdown files in both the Validate and Audit workflows.\n
Boundary markers: There are no explicit delimiters or instructions to ignore embedded prompts within the files being processed.\n
Capability inventory: The agent has access to Bash, Grep, and make tools, which could be exploited via successful injection.\n
Sanitization: No validation or sanitization of the markdown content is described before the agent interprets the data.

BuildAgent