speckit-specify
Warn
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local script at
.specify/scripts/bash/create-new-feature.shusing the user-provided feature description as a direct command-line argument. This creates a potential command injection vector if the input contains shell metacharacters such as semicolons, backticks, or subshell syntax. - [PROMPT_INJECTION]: The skill ingests untrusted natural language descriptions and interpolates them into shell commands and file templates. Ingestion point: User feature description; Boundary markers: None present in command string; Capability inventory: Shell script execution (
bash), git commands, and file writes (spec.md); Sanitization: Instructions for agent-side escaping of single quotes only. - [EXTERNAL_DOWNLOADS]: The skill performs
git fetch --all --pruneandgit ls-remoteto query the state of the remote repository and identify existing feature branches.
Audit Metadata