sdlc-edit-wiki

Warn

Audited by Gen Agent Trust Hub on Nov 13, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill identifies a "wiki repository" based on local file markers and executes a Python script (.wiki-llm/tools/wiki_llm.py) found within that repository. This pattern allows for the execution of arbitrary code provided by the target repository's environment.
  • [COMMAND_EXECUTION]: The skill makes extensive use of shell commands including git, grep, and python3 to interact with the file system and remote repositories.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: The skill reads and processes user-supplied wiki files from the wiki/ directory and schema definitions from .wiki-llm/schema/.
  • Boundary markers: The instructions do not define clear delimiters or "ignore" instructions when the agent processes or reformats content from the wiki.
  • Capability inventory: The agent possesses file write access, shell command execution capabilities, and the ability to perform network operations via git (fetch/push).
  • Sanitization: There is no evidence of escaping or sanitizing external wiki content before it is interpolated into log files or commit messages.
  • [DYNAMIC_EXECUTION]: The skill's primary logic for validation and drafting is offloaded to a script (wiki_llm.py) that is not bundled with the skill but is instead resolved and executed at runtime from the project directory.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 13, 2026, 09:20 AM