cwe-code-review
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted source code and pull requests as its primary input. While this creates a surface for indirect prompt injection, the skill's function is security auditing, and its instructions emphasize tracing code paths and confirming findings before assigning CWE identifiers to mitigate accidental obedience.
- [DYNAMIC_EXECUTION]: The instructions direct the agent to generate standalone Python scripts (
poc_<finding_slug>.py) to demonstrate confirmed vulnerabilities. This script generation is a core feature of the skill's intended purpose for security reporting. - [COMMAND_EXECUTION]: The skill utilizes local Python helper scripts (
scripts/cwe_lookup.pyandscripts/build_cwe_references.py) to manage and query the CWE catalog. These scripts use standard libraries to process local XML and JSON data and are intended for use within the review workflow.
Audit Metadata