The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted text from external signal sources to generate drafts, which presents an indirect prompt injection surface. Ingestion points: The FileSource adapter in core/adapters.py reads items from user-provided JSONL files. Boundary markers: Items are rendered as structured blocks in the prompt via core/generate.py but do not use explicit delimiters or 'ignore' instructions. Capability inventory: Includes local file-writing capabilities and LLM orchestration. Sanitization: Implements a multi-stage 'lint' process in core/lint.py that filters out fabrications and ensures every number or handle in the draft exists in the source text.\n- [EXTERNAL_DOWNLOADS]: The OpenRouterLLM implementation makes network requests to the OpenRouter API to perform completions. This is a well-known service used for the skill's primary function.\n- [COMMAND_EXECUTION]: The skill performs benign file system operations, such as reading input signal files and writing run logs, scored items, and final reports to a configurable data directory.

discover