shunk031-manage-public-private-skills

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill focuses on the administrative task of routing skill edits to the correct repository. It explicitly mandates security best practices, such as requiring the use of placeholders for credentials in committed evaluation prompts and preventing the disclosure of internal hostnames or service names in public repository metadata.
  • [INDIRECT_PROMPT_INJECTION]: As a tool for managing and editing other agent skills, this skill processes external instructions which constitutes an attack surface for indirect prompt injection.
  • Ingestion points: Untrusted content within SKILL.md files, scripts, or evaluation cases being added or modified via user requests (e.g., in evals/triggers.json and evals/evals.json).
  • Boundary markers: The instructions do not define specific delimiters to encapsulate or ignore instructions embedded in the skills being edited.
  • Capability inventory: The skill enables the agent to perform file system modifications (under ~/.agents/skills), execute git worktree commands, and perform dotfiles synchronization via chezmoi and make commands.
  • Sanitization: The skill proactively mitigates data exposure risks by requiring the use of placeholders for secrets and preventing the naming of private skills in public configuration files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 06:18 AM
Security Audit — agent-trust-hub — shunk031-manage-public-private-skills