The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple external sources including Taobao, Tmall, and Google Image search results.
Ingestion points: Data enters the context via taobao_fetch.py (HTML/JSON scraping), google_images.py (search result parsing), and scripts/main.py (CSV search results) as described in SKILL.md.
Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore' instructions when passing scraped content to the Gemini model in Phase 2.
Capability inventory: The skill has access to the filesystem (/tmp/store-teardown), network (requests, urllib), and can execute local scripts via uv run.
Sanitization: Absent. Scraped text (titles, descriptions, metadata) is directly interpolated into prompts for multi-modal analysis without escaping or filtering.
[EXTERNAL_DOWNLOADS]: The skill downloads images from various external domains such as alicdn.com and cdninstagram.com.
Evidence: scripts/download_images.py and scripts/google_images.py use urllib.request to fetch content from remote servers to the local filesystem.
[COMMAND_EXECUTION]: The skill uses uv run to execute local Python scripts which in turn perform network operations and filesystem writes.
Evidence: SKILL.md contains multiple shell blocks invoking uv run scripts/taobao_fetch.py, uv run scripts/download_images.py, etc.

store-teardown