update-docs
Warn
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill uses Node.js to dynamically load and parse project manifest files, which could lead to code execution if the files are malicious.
- Evidence: The script executes
node -p "require('./package.json').version ?? ''"to extract version information. - Impact: Because
require()in Node.js executes the target file if it contains JavaScript, a maliciouspackage.jsoncould lead to arbitrary code execution when the skill is run. - [DYNAMIC_EXECUTION]: The skill executes a Python script defined within its instructions at runtime.
- Evidence: The workflow pipes a multi-line Python script into
python3using a heredoc (<<'PYLINK') to verify internal documentation links. - Impact: While the script is visible in the skill definition, this pattern represents dynamic execution of code during the agent's session.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection by processing untrusted repository data.
- Ingestion points: The skill reads contents of all markdown files (
*.md), git logs, git diffs, and various project manifests (package.json,Cargo.toml,pyproject.toml,setup.py) to identify documentation drift. - Boundary markers: No explicit delimiters or safety instructions are provided to the agent to distinguish between its own instructions and the content of the files it is auditing.
- Capability inventory: The skill has broad capabilities including executing shell commands (
git,grep,find,node,python3,wc), writing to the file system, and committing changes to the repository. - Sanitization: There is no evidence of sanitization or filtering of the ingested content before it is processed by the LLM.
- [COMMAND_EXECUTION]: The skill executes multiple shell commands to inspect the repository environment, some of which interact with user-controlled file paths.
- Evidence: The skill uses
git,grep,find,sed,awk, anddateto audit roadmaps and evidence claims. - Context: While these are standard tools for a documentation audit skill, the execution occurs across a variety of repository files, including those that may be gitignored.
Audit Metadata