ci-health

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted data from the local filesystem and development tool outputs, creating an indirect prompt injection surface where external content is interpolated into the agent's context.
  • Ingestion points: Directory names, file sizes, git worktree lists, and security audit outputs from npm, pip, and cargo as seen in Phases 1-4 of SKILL.md.
  • Boundary markers: No explicit XML tags or delimiters are used to wrap the tool outputs or directory listings within the instructions.
  • Capability inventory: The skill can perform file deletions, prune git worktrees, and clean Docker images/builders when the --fix argument is passed.
  • Sanitization: There is no evidence of validation or sanitization of the filesystem metadata before it is presented in the final report.
  • [COMMAND_EXECUTION]: The skill uses system commands and development CLI tools (df, du, git, docker, npm, pip, cargo, dart) to perform audits and cleanup tasks. Destructive actions (Phase 5) are guarded by a required --fix argument and specific exclusion rules for system and active caches.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:58 AM
Security Audit — agent-trust-hub — ci-health