The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to manage the documentation infrastructure. Specifically, it utilizes mkdir -p for recursive directory creation across project layers (e.g., docs/01-项目层/, docs/02-技术层/) and includes basic bash conditional logic (if [ -d "$TARGET_DIR" ]) to safely initialize structures without overwriting existing data.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize external, untrusted project data stored in the "Raw Assets" layer, creating a potential vector for indirect prompt injection.
Ingestion points: Raw materials including meeting records, customer interview transcripts, web clips, and chat logs described in references/raw-layer.md.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" tags to prevent the agent from executing malicious prompts embedded within these raw inputs during the summarization or extraction process.
Capability inventory: The skill has the capability to write to the file system and execute shell commands for directory management, which could be abused if an injection is successful.
Sanitization: There are no explicitly defined sanitization or validation protocols for processing content extracted from raw materials before it is committed to the formal documentation layers.

project-knowledge-hierarchy