sh-style-guide

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to utilize external Python scripts (scripts/resolve_style_guides.py and scripts/validate_trigger_matrix_sync.py) for logic and validation. These are vendor-owned resources but are not provided in the skill package for security verification.
  • [DATA_EXFILTRATION]: A quick-start snippet (run_curl) demonstrates passing an API_TOKEN via a Bearer header to a variable URL. This pattern represents a potential for credential leakage if the URL variable is populated from untrusted or external sources.
  • [COMMAND_EXECUTION]: The guide includes examples using eval and dynamic command execution ($@) within a retry loop. While the skill advises caution for eval, these patterns are inherently risky if inputs are not strictly sanitized.
  • [PROMPT_INJECTION]: Indirect Prompt Injection Analysis: 1. Ingestion points: The skill processes user-provided .sh files, shebang lines, and CI workflow configurations. 2. Boundary markers: No delimiters are specified to distinguish between the skill's instructions and the untrusted shell code being analyzed. 3. Capability inventory: The skill encompasses capabilities for network requests (curl), file system manipulation (rm -rf, mktemp), and signal trapping. 4. Sanitization: The skill does not provide automated sanitization logic, instead relying on the instruction to 'Validate all external input before using it in commands'.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 08:38 PM