The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill performs an automated version check by fetching a configuration file from the author's GitHub repository at https://raw.githubusercontent.com/abpai/skills/main/versions.json.
[COMMAND_EXECUTION]: The skill instructions include logic to prompt the user to execute a shell command ('npx skills update distill') to perform a self-update.
[REMOTE_CODE_EXECUTION]: The self-update mechanism represents a remote code execution pathway, as it involves fetching remote version data and subsequently executing code from a package registry. This behavior is mediated by a user confirmation step.
[PROMPT_INJECTION]: The skill processes untrusted external data (such as codebases, research papers, and transcripts) using tools like Read, Glob, and Grep. This ingestion of external content without explicit boundary markers or sanitization creates a surface for indirect prompt injection.
Ingestion points: Files read during the Orient and Compress phases (SKILL.md).
Boundary markers: Not present; the skill lacks specific delimiters to separate ingested content from its internal logic.
Capability inventory: The skill can perform file writes (Write), invoke other agents (Agent), and suggest command-line operations (SKILL.md).
Sanitization: There is no evidence of filtering or escaping logic applied to the data before it is processed by the agent.

distill