The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from external, untrusted sources including X (Twitter), LinkedIn, Medium, and other platforms to generate blog drafts. This introduces a risk of indirect prompt injection where malicious instructions embedded in the source material could be processed and potentially followed by the agent.
Ingestion points: The workflow in SKILL.md and the cascading fallback chain in references/00-input-router.md describe fetching content from various social media and article URLs.
Boundary markers: The instructions do not explicitly provide boundary markers or directives for the agent to treat fetched data strictly as information rather than instructions.
Capability inventory: The skill instructions allow the agent to write files to the local filesystem (outputs/ directory) and interact with a remote MCP server for link building.
Sanitization: While the skill provides extensive rules for editorial "AI tell" removal and Taiwanization, it lacks specific sanitization logic for detecting or mitigating prompt injection patterns in external data.
[EXTERNAL_DOWNLOADS]: The skill utilizes an external link-building system via an MCP server located at https://index.tenten.dev/api/mcp. This involves sending the generated draft content to a remote vendor-controlled endpoint for link planning.
Vendor Context: The remote resource is owned and operated by the skill's author (tentenco) and is integrated as a core component of the content pipeline.

tenten-blog-writer