The Agent Skills Directory

[SAFE]: The skill consists of educational documentation, agent prompt templates, and a local shell script for linting. No malicious patterns such as prompt injection, data exfiltration, or unauthorized command execution were detected. The included validation script (scripts/validate-agent.sh) uses standard Unix utilities to check file structure without performing network operations or unsafe execution.
[INDIRECT_PROMPT_INJECTION]: The skill defines templates for agents that ingest and process external project files (e.g., AGENTS.md). This creates an ingestion surface for potentially untrusted data. However, this behavior is essential for the skill's purpose as a development tool. 1. Ingestion points: External project documentation files (AGENTS.md) and user-provided descriptions in templates. 2. Boundary markers: The skill encourages the use of structured JSON output and Markdown headers to separate instructions from data. 3. Capability inventory: The templates leverage standard development tools such as view_file, write_to_file, and run_command. 4. Sanitization: No explicit sanitization logic is provided in the templates, as they rely on the underlying AI platform's safety guardrails.

agent-development