The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation provides instructions to run several shell scripts and CLI commands for skill lifecycle management.
Evidence: Execution of scripts/create-agent-skill-repo.sh and scripts/skillport-sync.sh.
Evidence: Usage of npx skills add and npx skills update commands.
[REMOTE_CODE_EXECUTION]: The workflow is centered around fetching and installing executable skill packages from arbitrary remote GitHub repositories.
Evidence: npx skills add https://github.com/OWNER/REPO --list downloads content from a remote URL.
[PERSISTENCE]: The skill guides the user to establish automated background processes that persist across system restarts and sessions.
Evidence: Use of macOS LaunchAgents to run refresh tasks at login and wake-up.
Evidence: Use of Windows Task Scheduler to execute synchronization every 15 minutes.
[INDIRECT_PROMPT_INJECTION]: The distribution mechanism facilitates the ingestion of instructions from external repositories, which can introduce malicious content into the agent environment.
Ingestion points: Remote GitHub repositories via npx skills add commands.
Boundary markers: Absent for the ingested content in external repositories.
Capability inventory: Shell script execution, subprocess calls via npx, and file system writes to ~/.agents/skills.
Sanitization: Mentions credential and personal data scans for outbound pushes, but lacks explicit sanitization for incoming remote content.
[DYNAMIC_EXECUTION]: The refresh process involves invoking a script to dynamically regenerate agent instruction files.
Evidence: Invocation of scripts/regenerate-agents-md.py to update global guidance.

skillport-distribution-system