The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface for indirect prompt injection as it ingests untrusted data from Feishu APIs, including document metadata, wiki node titles, and user comments.
Ingestion points: Data is pulled into the agent context via commands like feishu-cli drive search, feishu-cli drive inspect, and feishu-cli comment list across various workflow files.
Boundary markers: There are no specific instructions or delimiters defined to isolate ingested external data from the agent's core instructions.
Capability inventory: The agent has significant write capabilities, such as modifying document permissions (perm add), deleting files (file delete), and managing wiki spaces (wiki delete-space).
Sanitization: The instructions do not specify any validation or sanitization routines for external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill is designed to execute shell commands using the feishu-cli utility to interact with the Feishu platform. It also utilizes jq and python3 for data manipulation.
Evidence: The allowed-tools field in SKILL.md explicitly permits the use of Bash(feishu-cli:*), Bash(jq:*), and Bash(python3:*).
[DATA_EXFILTRATION]: The skill provides functionality to move data between the local environment and the Feishu cloud service, which is its primary purpose.
Evidence: Workflow documentation in references/workflows/drive/workflow.md details commands for upload, download, pull, and push. The documentation highlights security features like SSRF protection that validates download URLs to prevent access to internal or loopback IP addresses.

feishu-cli-storage