The Agent Skills Directory

[SAFE]: The skill is a workflow manager for selecting and installing other agent skills. It operates primarily by reading a local catalog file (references/catalog.md) and proposing installation strings for the apm (Agent Package Manager) CLI.
[SAFE]: All external references point to the author's own GitHub organization (github.com/mizchi/*) or specific trusted sub-paths. No suspicious third-party or obfuscated URLs were found.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data (e.g., package.json, PR_BODY, or stated user intent) to suggest skills. However, it implements a mandatory human-in-the-loop confirmation step ('Confirm with the user before installing — propose, let them subtract') and uses a vetted catalog, which effectively mitigates the risk of automated malicious skill adoption.
[COMMAND_EXECUTION]: The skill documents the use of apm install and git commands. These are legitimate development tools required for its primary purpose (dependency management) and do not involve shell-piped remote execution or privilege escalation.
[DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic context injection (!command) syntax was found in SKILL.md or other files.
[OBFUSCATION]: Analysis of all markdown and YAML files confirms the absence of Base64, zero-width characters, or homoglyph attacks.

skill-selector