The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources, creating a significant attack surface for indirect prompt injection.
Ingestion points: Ingests content from Fathom meeting transcripts, LinkedIn posts/comments, TikTok trends, and Reddit threads via Apify scrapers.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded prompts within the retrieved external content.
Capability inventory: The skill has the ability to read local methodology files, execute external scraper tools (Apify), and write new entries to a Notion database.
Sanitization: Although the skill instructs the agent to anonymize client names in transcripts, it does not include logic to sanitize or filter out malicious instructions hidden in the scraped text.
[EXTERNAL_DOWNLOADS]: The skill relies on several external services to function.
Evidence: It retrieves data from Apify actors (e.g., apimaestro/linkedin-profile-posts, harvestapi/linkedin-post-search) and Fathom MCP tools. These are well-known services used as intended by the skill's description.
[DATA_EXPOSURE]: The skill accesses potentially sensitive information from meeting transcripts and internal work-streams.
Evidence: Processes Fathom meeting summaries and transcripts which may contain private client discussions. The skill attempts to mitigate this by mandating anonymization of names and NDA-sensitive details before storing information in Notion.

content-research