codewiki-generator
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/codewiki_analyze.pyexecutes several system commands usingsubprocess.runto extract metadata from the target repository. - It invokes
go list,cargo metadata,swift package, andctagswithin the repository directory to identify dependencies and symbols. - [EXTERNAL_DOWNLOADS]: The skill depends on external Node.js packages and CLIs to build and deploy the documentation site.
- The workflow in
SKILL.mdtriggersnpm installandnpxto manage VitePress and its plugins. - The deployment instructions in
references/deploy-cloudflare.mdrecommend installing thewranglerCLI from the npm registry. - [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it interprets and processes content from the analyzed codebase.
- Ingestion points:
scripts/codewiki_analyze.pyreads source code,package.json,README.md, and other text files from the target repository. - Boundary markers: While generated documentation includes
# Related Codemarkers, there are no explicit instructions to the LLM to ignore embedded commands within the analyzed source data. - Capability inventory: The skill can perform file system writes and execute subprocesses (npm, npx, wrangler) based on the documentation plan.
- Sanitization: There is no sanitization or filtering of the repository's content before it is processed by the agent to generate documentation.
Audit Metadata