The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads multiple CSV data files and news summaries from a remote server that is not a recognized trusted vendor or service.
Evidence: scripts/update-data.sh uses curl to fetch files from https://zxuuu.link/longterm-agu/data.
[COMMAND_EXECUTION]: The skill instructions require the agent to execute shell scripts directly to update and verify data, which increases the attack surface for local command injection or environment manipulation.
Evidence: SKILL.md mandates running scripts/update-data.sh and scripts/check-data.sh.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes news data summaries from an untrusted external source without sanitization or strict boundary markers.
Ingestion points: assets/data/news_breakfast.csv (read in the 'Step 4: 消息面分析' section of SKILL.md).
Boundary markers: None identified in the skill instructions to prevent the agent from following instructions embedded in the news data.
Capability inventory: The skill has the ability to execute shell scripts and write files to the ./reports/ directory.
Sanitization: No sanitization or validation of the downloaded news content is performed before processing.

investment-analysis