skills/caidanw/skills/update-readme/Gen Agent Trust Hub

update-readme

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local bash script (update-readme/update-readme.sh) to automate the regeneration of the 'Available Skills' table in README.md.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) by aggregating data from external skill files into the primary documentation.
  • Ingestion points: The script parses the name and description fields from the YAML frontmatter of all SKILL.md files within the repository subdirectories.
  • Boundary markers: There are no markers or instructions to isolate the parsed content or prevent the agent from obeying instructions embedded in those fields.
  • Capability inventory: The script performs file-system reads across the repository and writes the compiled table to the root README.md (update-readme.sh, line 102).
  • Sanitization: The script performs minimal cleaning (stripping quotes and spaces) but does not escape Markdown syntax (like | or backticks), allowing potentially malicious metadata to corrupt the table structure or inject content into the main repository README.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 06:15 AM
Security Audit — agent-trust-hub — update-readme