The Agent Skills Directory

[SAFE]: The skill establishes a clear documentation and escalation workflow using markdown templates and local file storage.
[PROMPT_INJECTION]: The skill processes untrusted data via the issue_description parameter, creating a surface for indirect prompt injection in the generated reports.
Ingestion points: The issue_description input in SKILL.md is directly interpolated into report files.
Boundary markers: The templates do not use specific delimiters to isolate the description from the rest of the report context.
Capability inventory: The skill is limited to file system operations via Read, Write, Edit, Grep, and Glob tools.
Sanitization: No automated sanitization of input strings is performed before they are written to disk.
[SAFE]: The Python code block provided in the 'Step 3: Determine Routing' section is intended for logical illustration in documentation and is not executed by the agent.
[SAFE]: References to security-sensitive findings and example credentials (e.g., sk_live_abc123) are used strictly as illustrative examples for reporting purposes.

escalate-issues