code-analyzer
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a local shell script (scripts/scan-codebase.sh) to perform codebase scanning. The script utilizes standard system tools (find, grep, git, sed) to aggregate file structure and metadata. Evidence: The SKILL.md instructions explicitly call for running the bash script.- [INDIRECT_PROMPT_INJECTION]: The skill extracts text from project files using specific markers (AICODE-NOTE, AICODE-TODO, AICODE-FIX), creating an attack surface for indirect prompt injection. Maliciously crafted comments in the scanned codebase could attempt to override agent behavior during the interpretation phase.
- Ingestion points: Source code files processed by the extract_markers function in scripts/scan-codebase.sh.
- Boundary markers: Absent. The skill does not instruct the agent to treat extracted marker text as untrusted or provide clear delimiters to separate it from system instructions.
- Capability inventory: The skill possesses the ability to execute shell scripts and instructs the agent to read and follow execution paths through the codebase.
- Sanitization: The script implements json_escape to ensure the output remains valid JSON, but it does not perform semantic sanitization or validation of the extracted text content.
Audit Metadata