The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted source code and configuration files, creating a surface for injection.
Ingestion points: Reads local source code, container manifests, and CI/CD configurations.
Boundary markers: Lacks technical delimiters to isolate untrusted code from instructions.
Capability inventory: Authorized to use Bash for running security tools and Read for accessing files.
Sanitization: No explicit filtering of content read from analyzed files.
[COMMAND_EXECUTION]: Provides templates for executing network and security tools via Bash, including nmap, sqlmap, and dig.
[DATA_EXFILTRATION]: Outlines methods for discovering sensitive data, such as auditing shell history files and searching for credentials in user directories.
[EXTERNAL_DOWNLOADS]: Recommends installing security tools from official package registries like NPM and PyPI, as well as GitHub repositories.
[PRIVILEGE_ESCALATION]: Provides instructions for auditing system privileges, including checking sudo permissions and SUID binaries during security reviews.

security-reviewer