trace

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes several standard command-line utilities including ghq, gh (GitHub CLI), find, and date. These are used to manage local repositories, query GitHub issues/PRs, and timestamp trace logs according to the stated purpose of the skill.
  • [EXTERNAL_DOWNLOADS]: When the --repo [url] flag is used, the skill invokes ghq get to clone external repositories from GitHub. This is a legitimate function for a project discovery tool and targets a well-known service.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing data from git history, file contents, and session logs.
  • Ingestion points: SKILL.md (Agent 1/2) searches file contents and git history; scripts/dig.py reads Claude Code session logs from .jsonl files.
  • Boundary markers: Not explicitly implemented in the instructions to demarcate ingested content from agent instructions.
  • Capability inventory: The skill has the capability to create directories, execute local tools (ghq, gh), and interact with the Oracle MCP for logging.
  • Sanitization: scripts/dig.py applies basic sanitization by truncating user message snippets to 80 characters.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 02:36 AM