The Agent Skills Directory

[SAFE]: The skill references established security frameworks and documentation from trusted organizations such as MITRE, FIRST, and NIST. All external links point to official and well-known security domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (scanner outputs, system banners, plugin descriptions) which presents an attack surface for indirect prompt injection. 1. Ingestion points: Scanner result files, vulnerability plugin descriptions, and system banners processed during tuning. 2. Boundary markers: Present. The skill contains a 'Prompt Injection Safety Notice' explicitly instructing the agent to disregard instructions embedded in these data sources. 3. Capability inventory: The skill is restricted to file system read operations (Read, Grep, Glob) and lacks network or code execution capabilities. 4. Sanitization: The skill relies on instructional guardrails and manual verification workflows rather than programmatic sanitization.
[SAFE]: No hardcoded credentials, malicious persistence mechanisms, or unauthorized privilege escalation commands were found. The tool permissions requested are appropriate for the task.

scanner-tuning