The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to search for and read sensitive information, such as hardcoded credentials, API keys, and authentication logic, as a core part of its security audit functionality. This behavior is restricted to local analysis for report generation.
[COMMAND_EXECUTION]: Employs shell commands via bash to interact with the project repository, using git to identify changes in pull requests and grep to scan for vulnerability patterns in code files.
[PROMPT_INJECTION]: The skill processes untrusted data by reading source code files from the repository. This creates a surface for indirect prompt injection, as malicious instructions embedded within the analyzed code (e.g., in comments or string literals) could theoretically influence the agent's behavior during the audit process.
Ingestion points: Reads PHP, JavaScript, TypeScript, and SQL files from the local filesystem.
Boundary markers: None explicitly defined in the task prompt for the security-specialist sub-agent.
Capability inventory: File system access (Read/Write), shell command execution (git/grep), and task delegation to other agents.
Sanitization: No specific sanitization or filtering of the ingested code content is described.

security-audit