create-verification-skill
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and executes shell commands discovered in untrusted repository files, such as
package.json,Makefile, and CI configurations. This creates a surface where a malicious repository can define harmful scripts that the agent will execute during the command discovery phase. - Ingestion points:
package.json,Makefile, CI configuration files. - Capability inventory: Shell command execution via discovered project scripts.
- Boundary markers: None specified to differentiate between intended commands and malicious injections.
- Sanitization: No evidence of command validation or sanitization before execution.
- [COMMAND_EXECUTION]: The workflow explicitly instructs the agent to "find the command... and run it," leading to the direct execution of arbitrary shell commands provided by the local workspace environment.
- [DYNAMIC_EXECUTION]: The skill involves multiple layers of dynamic execution: it generates a new functional skill file (
.claude/skills/verify/SKILL.md), instructs the agent to execute that generated workflow immediately, and calls functions from a local JavaScript file (dist/features/verification-ledger/index.js) to manage its state.
Audit Metadata