The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to dynamically fetch its authoritative instructions and formatting standards from a remote URL (https://agentskills.io/llms.txt) at runtime.
Ingestion points: Fetches llms.txt and subsequent Markdown pages from the agentskills.io domain into the agent context.
Boundary markers: Absent. The skill instructions state that the fetched remote content is the 'only fact standard' and takes priority over local instructions.
Capability inventory: The skill can execute shell commands via uv, pi, and skills-ref, and can read/write files in the local workspace.
Sanitization: No validation or sanitization is performed on the remote instructional content before it is adopted as the agent's operating standard.
[EXTERNAL_DOWNLOADS]: The skill fetches instructional Markdown files from agentskills.io. It also utilizes uvx to download and run the skills-ref package for validation, which involves fetching code from a remote registry.
[COMMAND_EXECUTION]: The skill executes various CLI tools to perform its development and evaluation tasks.
The scripts/run_trigger_eval.py script uses subprocess.run to execute the pi CLI tool with user-supplied queries.
The SKILL.md file instructs the agent to use uvx skills-ref for validating the formatting of new skills.
It uses uv run to execute the local evaluation script with specific environment requirements.

dbk-skill-dev