The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. 1. Ingestion points: The skill ingests pending decisions from conversation context and local documents (Phase 1, SKILL.md). 2. Boundary markers: Structural delimiters like '[E01]' are used to index items, but no instructions are provided to sanitize or ignore embedded prompts. 3. Capability inventory: The agent is instructed to write decided items back to source documents/TODO files (Phase 4, SKILL.md) and generate HTML forms via a generator script. 4. Sanitization: There are no instructions for sanitizing, escaping, or validating the input data before processing or persistence.
[DYNAMIC_EXECUTION]: The skill workflow involves using a Python script (scripts/formular/build_form.py) from the decision-draw skill to generate a self-contained HTML form from JSON data at runtime. This creation of executable content (HTML/JS) from dynamic inputs represents a dynamic execution pattern for user interface generation.

decision-briefing