The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and synthesize external data (documentation, repositories, and web content) into executable agent instructions, which creates a surface for indirect prompt injection. This surface is documented in 'references/synthesis-path.md'. The skill includes mitigations by explicitly instructing the agent to treat external content as untrusted data and apply trust and safety rules during the ingestion process.
[COMMAND_EXECUTION]: The documentation in 'EVAL.md' and 'references/registration-validation.md' describes workflows that involve local command execution using tools like 'codex exec' and 'uv run'. These are documented as developer-controlled steps for testing and validating generated skills.
[EXTERNAL_DOWNLOADS]: The skill references external specifications and upstream repositories (e.g., 'github.com/anthropics/skills', 'agentskills.io') for synchronization and provenance tracking. These references target well-known and trusted technology organizations.

skill-writer