The Agent Skills Directory

[DYNAMIC_EXECUTION]: In scripts/baoyu_image_gen_core.ts, the skill dynamically assembles a Python script as a string and executes it via python3 -c. This mechanism is used to export configuration variables from wechat-publisher.yaml into the Bun execution environment.
[COMMAND_EXECUTION]: Several scripts utilize subprocess.run or Bun.spawnSync to execute external processes. For instance, scripts/multi_publish.py calls the wechatsync CLI, and scripts/generate_image.py invokes bun to run image generation logic. While arguments are passed as lists, this orchestration of multiple environments increases the complexity of the execution chain.
[EXTERNAL_DOWNLOADS]: The scripts/image_handler.py component contains logic to download images from arbitrary user-provided or web-sourced URLs using the requests library. This is part of the article processing workflow but involves outbound requests to non-whitelisted domains.
[INDIRECT_PROMPT_INJECTION]: The skill defines a research phase that ingests data from external community sites like Reddit, HN, and V2EX. This material is then used as context for the agent to write articles.
Ingestion points: Untrusted text from community forums and metadata from external images (in scripts/image_handler.py).
Boundary markers: Absent. There are no clear delimiters or instructions to the agent to disregard commands embedded within the gathered research material.
Capability inventory: High. The skill can execute shell commands, perform network operations, and read/write local files.
Sanitization: Limited. While scripts/html_converter.py performs basic HTML escaping and protocol filtering for links, there is no comprehensive sanitization of the research data entered into the agent's context.

skill-wechat-publisher