knowhow-run

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses standard shell utilities including ls, grep, and find to verify directory existence and search for markdown workflow files within the local workspace. These operations are scoped to the .knowhow/ directory for legitimate discovery purposes.
  • [PROMPT_INJECTION]: The skill contains a directive to prioritize itself over built-in tools ("BẮT BUỘC gọi skill này, KHÔNG phải skill 'run' built-in"). This is an instruction to ensure specific project workflows are followed instead of default agent behavior.
  • [PROMPT_INJECTION]: The skill is designed to ingest and execute instructions from local markdown files, establishing an indirect prompt injection surface.
  • Ingestion points: Files located in .knowhow/skills/ and .knowhow/workflows/.
  • Boundary markers: None specified; the agent is instructed to read the entire content and follow it.
  • Capability inventory: Shell utilities (ls, find, grep) and file write operations for logging to wiki/log.md and .knowhow/schema-signals.md.
  • Sanitization: No sanitization of the markdown file content is described before the agent follows the instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 07:02 AM
Security Audit — agent-trust-hub — knowhow-run