sdlc-build-task

Pass

Audited by Gen Agent Trust Hub on Nov 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell commands including Git for branch and worktree management, and executes a local project script (.wiki-llm/tools/wiki_llm.py) to validate wiki consistency. These operations are essential to the development workflow and are performed on local repositories identified by the user.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external Markdown files within a project wiki repository. \n
  • Ingestion points: Task files, Design contracts, and convention documents (SKILL.md, read-protocol.md). \n
  • Boundary markers: No specific delimiters are used to wrap ingested data. \n
  • Capability inventory: The agent can write files, create git worktrees/branches, and execute local scripts. \n
  • Sanitization: No programmatic sanitization is applied; however, the skill enforces three manual approval gates (Intent, Design, Change-set) to prevent the automated execution of malicious instructions embedded in wiki data.\n- [DATA_EXFILTRATION]: No unauthorized network operations or data exfiltration patterns were detected. The skill includes explicit instructions for the AI to avoid copying sensitive data or secrets into summary reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 13, 2026, 09:20 AM