The Agent Skills Directory

[SAFE]: The skill functions as a security review framework, providing detailed checklists for identifying common vulnerabilities like XSS, SQL injection, and mobile security gaps.
[PROMPT_INJECTION]: No malicious instructions or attempts to bypass safety filters were found. The skill uses natural instructional language to scope its auditing functions.
[DATA_EXFILTRATION]: The skill does not perform network operations or unauthorized file access. References to local skill paths (e.g., ~/.agents/skills/) and vendor plugins are used for legitimate integration within the agent environment.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to ingest and analyze external source code, creating an attack surface for indirect prompt injection, it focuses on reporting patterns rather than executing untrusted content.
Ingestion points: Reads code using Read, Grep, and Glob tools.
Boundary markers: Relies on agent-level handling of file content.
Capability inventory: Uses standard analysis tools (Read, Bash, Glob, Grep) without risky network or administrative privileges.
Sanitization: Focuses on identifying missing sanitization in target code as part of its audit process.

stack-security-coder