The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill inherently processes data from external sources, creating a surface for potential indirect prompt injection attacks.
Ingestion points: The skill reads external data from Lark Bitable records via commands like +record-list and +record-search, and metadata from external URLs via +url-resolve (SKILL.md).
Boundary markers: The skill suggests using structured NDJSON artifacts for data storage and processing, providing a level of structural separation (SKILL.md).
Capability inventory: The agent has the capability to execute lark-cli subcommands, as well as python3 scripts and jq filters for data analysis on the local environment (SKILL.md).
Sanitization: The instructions do not explicitly mandate prompt-level sanitization or escaping of ingested record content before it is re-evaluated by the agent.
[COMMAND_EXECUTION]: The skill provides detailed instructions and examples for the agent to execute shell commands. This includes project-specific tools like lark-cli as well as general-purpose utilities like python3 and jq for data processing and analysis. These are legitimate capabilities required for the skill's primary purpose of Bitable data management and automation (SKILL.md).

lark-base