The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface for indirect prompt injection due to its core function of processing untrusted external data.
Ingestion points: According to the description and Step 1, the skill pulls "live data from the client's outreach platform" (LinkedIn/Email) and processes user-provided exports such as "CSV / Sheet / paste" content.
Boundary markers: The instructions in SKILL.md lack explicit delimiters or "ignore embedded instructions" warnings for the data being processed. Malicious instructions embedded in a LinkedIn reply or a CSV cell could potentially influence the agent's behavior.
Capability inventory: The skill is capable of performing network operations to fetch data, performing "reply analysis" (natural language processing), and writing new pages to a "client's reporting area."
Sanitization: There is no requirement or instruction to sanitize, escape, or validate the content of prospect threads before they are quoted or used to generate "ready-to-send" drafts, which could lead to the agent echoing malicious payloads or following injected commands.

weekly-outreach-report