skills/zixun-github/aisdlc/spec-init/Gen Agent Trust Hub

spec-init

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill performs local system operations using git and shell utilities to manage repository state and file structures.
  • Evidence: Both spec-create-branch.sh and spec-create-branch.ps1 execute git checkout -b, git fetch, and file system commands like mkdir and rm to initialize the spec environment.- [PROMPT_INJECTION]: The skill processes untrusted requirement text provided by users, creating an indirect prompt injection surface.
  • Ingestion points: User-supplied text is written to a temporary file (_sdlc-raw-temp.md) or read from a user-provided path via the $SourceFilePath parameter in scripts/spec-create-branch.ps1 and scripts/spec-create-branch.sh.
  • Boundary markers: Absent; raw text is processed without delimiters or instructions to ignore embedded instructions.
  • Capability inventory: The skill has permissions to create/delete files and manage git branches within the repository.
  • Sanitization: Branch names and directory names are validated against a strict kebab-case regex in the bash implementation, but the content of the requirement text itself is written to requirements/raw.md without sanitization or filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 02:43 AM