The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from several ingestion points, creating a risk that malicious instructions embedded in external content could influence agent behavior.
Ingestion points: Product URLs, attached product images, and live web search results across social media platforms (Instagram, TikTok, YouTube) as described in SKILL.md (Stage 1) and publish-and-report-workflow.md (Stage 4).
Boundary markers: The skill does not define explicit delimiters or instructions to the agent to disregard potential commands found within the fetched external data.
Capability inventory: The agent has the capability to publish content to Meta Ads and initiate video generation tasks in Higgsfield Marketing Studio, which involves credit consumption and financial expenditure.
Sanitization: The skill lacks documented sanitization or filtering logic for the content retrieved from external niche-related websites.
[EXTERNAL_DOWNLOADS]: The skill performs automated data retrieval from external sources to conduct niche research.
Evidence: SKILL.md specifies that the agent will "run trend research — a spread of web searches... plus at least two source-page fetches" to synthesize content briefs.

higgsfield-content-factory