The Agent Skills Directory

[SAFE]: The skill is composed of markdown instructions and reference files for prompt engineering. It does not include any scripts, binaries, or automated network operations.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through the ingestion of external documents and links. This is a common characteristic of document-analysis tools. The risk is minimized by the skill's structure: it requires a mandatory clarification phase (Step 2) with 5-10 user questions before outputting any results, preventing the direct and automated execution of instructions found in untrusted data. 1. Ingestion points: SKILL.md Step 1 (Links, Documents). 2. Boundary markers: Absent. 3. Capability inventory: No internal scripts; uses platform-provided browsing/reading tools. 4. Sanitization: Absent.

prompt-architect