The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from user notes in the "05_每日记录" folder and system reports in "memory/network-health/". This creates an architectural surface for indirect prompt injection, as malicious instructions embedded within user notes could potentially influence the agent's behavior during automated maintenance tasks. This risk is fundamental to the skill's purpose of analyzing and organizing text content.
Ingestion points: Reads user-generated content from 05_每日记录/ and automated health reports from memory/network-health/.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" markers when processing the content of ingested notes.
Capability inventory: The skill can write processing logs and updated indexes to the memory/ directory and execute a local Python script via the shell.
Sanitization: No specific validation or sanitization steps for note content are described.
[COMMAND_EXECUTION]: The skill documentation includes an optional automation workflow that executes a local Python script: python3 ".agents/skills/index-note/scripts/batch_network_integration.py". This is a documented extension of the skill's functionality used for idempotent network updates and operates on local file paths.

index-note