claude-code-commands
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill generates templates for Claude Code slash commands that interpolate untrusted user input ($ARGUMENTS) directly into shell command blocks, which could lead to command injection.
- Ingestion points: Untrusted data enters via the
$ARGUMENTSvariable within the command templates defined inreferences/patterns.md(e.g., the Issue-Linked Command and DB Command templates). - Boundary markers: There are no explicit boundary markers or instructions to ignore embedded shell meta-characters within the
$ARGUMENTSstring in the templates. - Capability inventory: The templates include capabilities for executing subprocesses via
bash, specifically involvinggit,gh(GitHub CLI),grep, andnpmcommands inreferences/patterns.md. - Sanitization: The skill does not provide or recommend any sanitization, escaping, or validation logic for the
$ARGUMENTSbefore they are used in shell commands, althoughreferences/validations.mdcheck for the literal syntax of the variable itself.
Audit Metadata