security-watch
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a security linter that executes during file write operations to identify hardcoded secrets (AWS, OpenAI, GitHub tokens, etc.), dangerous functions (eval, innerHTML), and SQL injection patterns.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data (file contents) through the PreToolUse hook to perform its scanning.
- Ingestion points: The scripts/scan-security.sh script reads content from the Write and Edit tool calls via command line arguments or standard input.
- Boundary markers: The content is processed as a raw string without specific delimiters or instructions to the scanner to ignore embedded commands.
- Capability inventory: The script can block the agent's write/edit actions by returning a non-zero exit code if issues are found.
- Sanitization: The script performs read-only regex matching without executing the input content, minimizing the risk of the agent being influenced by instructions embedded in the scanned files.
- [COMMAND_EXECUTION]: The processing script uses standard shell commands (grep, echo) to evaluate text. Content is handled through variable piping and standard input, which does not introduce command injection risks in the context of the scanner's execution.
Audit Metadata