The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection where malicious instructions hidden in project files could be processed and followed by the agent.
Ingestion points: Phase 0 in SKILL.md identifies and reads existing design or requirements documents from directories such as docs/ or prd/ or from user-specified file paths.
Boundary markers: The instructions do not specify any delimiters or 'ignore-instructions' warnings to isolate the imported context from the agent's core logic.
Capability inventory: The skill possesses the capability to write files to the local file system (PRD-*.md) and execute shell commands via the lark-cli tool to create or update remote documents.
Sanitization: No sanitization, validation, or escaping is performed on the data extracted from the local project files before it is incorporated into the generated PRD.
[COMMAND_EXECUTION]: The skill uses the lark-cli command-line utility to interact with the Feishu/Lark platform. It constructs and executes shell commands such as lark-cli docs +create and lark-cli docs +update using content derived from user input and ingested files.

create-prd