agent-readiness

Warn

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a high number of shell commands to inspect repositories, including git, gh, glab, and several package manager CLIs (npm, yarn, pip, cargo, etc.). These tools interact with the local filesystem and remote platform APIs.
  • [EXTERNAL_DOWNLOADS]: In Phase 1, the skill clones external repositories into the /tmp directory based on URLs provided in the user arguments.
  • [REMOTE_CODE_EXECUTION]: The audit process evaluates test executability by running dry-run commands such as pytest --collect-only or jest --listTests. These commands typically load the target repository's source and configuration files (like conftest.py), which can lead to arbitrary code execution if the audited repository is malicious.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from the contents of the audited repositories.
  • Ingestion points: Clones and reads arbitrary codebases provided via the $ARGUMENTS variable in SKILL.md.
  • Boundary markers: Absent. The skill does not use specific markers to isolate untrusted repository content from its execution logic.
  • Capability inventory: Extensive subprocess execution (git, gh, npm) and local file writes (JSON report generation) across all phases.
  • Sanitization: Absent. Data extracted from the audited repository is used directly to calculate scores and generate the final report.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 23, 2026, 02:25 PM
Security Audit — agent-trust-hub — agent-readiness