ponytail
Pass
Audited by Gen Agent Trust Hub on Jul 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a shell script
scripts/install.shdesigned to detect the agent runtime (e.g., Claude Code, Codex, Gemini) and execute the corresponding plugin installation commands. This execution is conditional and requires theAPPLY=1environment variable to be set by the user. - [EXTERNAL_DOWNLOADS]: The installation instructions and script reference external repositories for plugin data, specifically
github.com/DietrichGebert/ponytail(upstream) andgithub.com/akillness/oh-my-skills(catalog). It also utilizesnpxto fetch theskillsCLI tool during setup. - [PROMPT_INJECTION]: The skill contains defensive instructions in the 'Edge Cases & Pitfalls' section, explicitly ordering the agent to never cut code related to trust-boundary validation, data-loss handling, security (authn/authz, secrets, injection defenses), or accessibility.
- [DATA_EXFILTRATION]: The skill uses
Grep,Glob, andReadtools to audit the repository for tech debt markers (ponytail:) and over-engineering patterns. While this involves broad filesystem access, no patterns of external data exfiltration were detected.
Audit Metadata