The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive automation by executing local scripts via shell commands.
Executes topic discovery scripts (search.js) within the ~/.openclaw/skills/brave-search directory.
Runs Python-based image generation tools (generate.py) located in ~/clawd/projects/MediaClaw/skills/code-material-gen/scripts/.
Invokes a publication script (publish.py) that handles the automated upload to WeChat Official Accounts.
[DATA_EXFILTRATION]: The skill accesses sensitive identity and session information to perform its tasks.
Reads user persona data and Daniel's specific profile from ~/.openclaw/workspace-content/USER.md.
Explicitly mentions a dependency on "微信MP cookie" (WeChat MP cookies) for its automated browser-based publishing workflow, which presents a risk of credential exposure if the agent's logic is subverted.
[PROMPT_INJECTION]: The workflow is vulnerable to indirect prompt injection through search engine results.
Ingestion points: The skill fetches and processes real-time data from brave-search and Perplexity to generate article topics and competitive analysis.
Boundary markers: There are no explicit markers or instructions to treat search results as untrusted data, increasing the risk that embedded malicious instructions in web content could influence the agent.
Capability inventory: The skill possesses significant capabilities, including shell execution for multiple scripts, local file system modification (creating directories and saving articles), and network publication.
Sanitization: The skill lacks mechanisms to sanitize or filter search results before they are interpolated into the content generation prompts.

daily-gzh-content