The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill employs a Python utility (scripts/refresh_sources.py) to synchronize documentation and presentation examples from docs.deckset.com and api.github.com. These targets are official service domains or whitelisted repositories, making the downloads safe under trust guidelines.
[DYNAMIC_EXECUTION]: The workflow requires the execution of a local Python script (refresh_sources.py) to manage reference materials. The script uses standard libraries for networking and filesystem operations, including atomic writes and file cleanup, which follow safe development practices.
[COMMAND_EXECUTION]: The skill instructions direct the agent to run shell commands (e.g., uv run scripts/refresh_sources.py) to maintain its knowledge base. These commands are scoped to the skill's local directory and intended for maintenance tasks.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and transform untrusted conversation context into structured markdown, which introduces a surface for indirect prompt injection.
Ingestion points: User conversation history and requests (extracted in SKILL.md Step 3).
Boundary markers: The instructions lack specific delimiters or escaping rules to separate user-provided content from the agent's instructions.
Capability inventory: The skill has the ability to execute local scripts with network access and filesystem write/delete permissions.
Sanitization: No explicit sanitization or filtering of the ingested conversation data is documented in the instruction set.

deckset