The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation includes instructions to execute local bash commands for environment activation and to run data retrieval scripts (e.g., views_runner.py) from a sibling directory (findata-toolkit-cn). These are intended operations for local data processing and environment management.
[EXTERNAL_DOWNLOADS]: The skill uses pip to install dependencies from a local requirements file, which is a standard procedure for managing Python-based tool environments.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface due to its core function of ingesting and analyzing unstructured market sentiment data from external social media and news platforms.
Ingestion points: The workflow identifies 'Media/Sentiment' and social media platforms like 'Xueqiu' and 'Guba' as primary data inputs in SKILL.md and references/gap-analysis-methodology.md.
Boundary markers: The provided methodology and templates do not include specific delimiters or instructions to ignore embedded prompts within the ingested market data.
Capability inventory: The skill is capable of executing local scripts and performing network requests through its associated data toolkit.
Sanitization: The documentation does not describe explicit sanitization, validation, or filtering of external content, relying instead on the underlying agent's standard safety protocols.

sentiment-reality-gap