Command Development
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill teaches the use of bash execution syntax (
!\command`) to gather dynamic context such as repository status or environment variables. It explicitly guides developers to use theallowed-toolsfrontmatter field to restrict bash execution to specific command patterns (e.g.,Bash(git:*)), significantly reducing the attack surface.\n- **[DATA_EXFILTRATION]:** The documentation describes the@syntax for including file contents in prompts. To prevent unauthorized file access, it provides 'Validation Patterns' that demonstrate how to verify file existence and validate user-supplied paths before they are processed by the agent.\n- **[PROMPT_INJECTION]:** The skill provides instructions on ingesting user-supplied data via positional arguments ($1, $2) and the$ARGUMENTS` variable.\n - Ingestion points: Untrusted data enters the agent context through CLI arguments defined in
SKILL.mdandfrontmatter-reference.md.\n - Boundary markers: The skill suggests using clear instructional headers and framing to separate data from instructions.\n
- Capability inventory: Examples demonstrate potential subprocess execution (git, npm, gh), file-reading, and file-writing tools across the
examples/andreferences/directories.\n - Sanitization: The skill contains a dedicated 'Validation Patterns' section in
SKILL.mdandplugin-features-reference.mdthat teaches developers how to perform input validation and existence checks to ensure that untrusted data does not lead to unintended command execution.
Audit Metadata