The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Weibo timelines, posts, and search results which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Data is ingested through tools like timeline.home.list, post.get, post.replies.list, and search.weibo as described in SKILL.md.
Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore' instructions when processing the content of posts or search results.
Capability inventory: The skill possesses significant write capabilities, including post.create, comment.create, and article.publishMarkdown, which could be abused if an injection is successful.
Sanitization: While the skill emphasizes parsing JSON output (.ok == true), there is no evidence of sanitization or filtering of the natural language content retrieved from the web.
[COMMAND_EXECUTION]: The skill relies on several local CLI tools and shell scripts to perform its tasks.
Evidence: SKILL.md and references/usage-patterns.md describe the use of weibo-webmcp-cli, uxc, and npx. The scripts/ensure-links.sh script executes a sibling script skills/webmcp-bridge/scripts/ensure-links.sh. The scripts/validate.sh script also uses rg (ripgrep) to check file patterns.
[EXTERNAL_DOWNLOADS]: The skill requires downloading browser binaries for its operation.
Evidence: SKILL.md instructs the user/agent to run npx playwright install to set up the necessary environment for web automation.

weibo-webmcp