The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves ingesting and analyzing untrusted content from the user, including text, URLs, or file paths, to generate a visual infographic. Malicious instructions embedded in this source material could potentially influence the agent's analysis or prompt generation logic.
Ingestion points: User-supplied content is processed in SKILL.md (Step 1) and stored locally in source.md.
Capability inventory: The skill utilizes write_file for creating analysis and prompt files, and image_generate for final output.
Boundary markers: The skill employs Markdown templates (references/structured-content-template.md) to organize data, but it lacks specific security delimiters to prevent the LLM from accidentally executing embedded instructions found in the user content.
Sanitization: The instructions explicitly direct the agent to 'strip any credentials, API keys, tokens, or secrets before including in outputs,' which is a proactive measure against data exposure, though it does not fully prevent behavioral injection.
[NO_CODE]: The skill consists entirely of Markdown instructions and reference documentation. It does not include executable scripts, shell commands, or external package dependencies, which significantly reduces its technical attack surface.

baoyu-infographic