The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes instructions for the agent to execute shell commands to manage networking infrastructure. Specifically, it provides commands to terminate existing SSH processes (kill -9 $(pgrep -f "ssh.*id_vultr.*8888")) and initiate a background SSH tunnel (ssh -f) to an external IP address (45.76.185.1).
[CREDENTIALS_UNSAFE]: The proxy recovery instructions reference a specific sensitive file path for an SSH private key (/home/ubuntu/.ssh/id_vultr). Hardcoding specific credential paths and external server IPs in skill instructions facilitates unauthorized remote access and credential exposure.
[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves using a browser tool to download and parse PDF announcements from external sources (CNINFO). This ingestion of untrusted data from the web represents an attack surface where a maliciously crafted PDF could contain instructions designed to override agent behavior during the parsing and database update process.
Ingestion points: PDF files downloaded via the browser tool from www.cninfo.com.cn and parsed using pymupdf (referenced in SKILL.md).
Boundary markers: None identified in the provided instructions or data schemas to separate parsed content from system instructions.
Capability inventory: The skill has the ability to execute shell commands and Python scripts (scripts/update_db.py, scripts/export_excel_v3.py).
Sanitization: No sanitization or validation logic is mentioned for the text extracted from PDFs before it is processed by the update scripts.

lithium-battery-db