The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external websites.
Ingestion points: Content is scraped from user-provided company URLs using tools like review-scraper, web-archive-scraper, and linkedin-post-research in SKILL.md (Phase 1 and Phase 3).
Boundary markers: There are no explicit instructions or delimiters defined to prevent the model from following malicious instructions that might be embedded in the scraped web content.
Capability inventory: The skill has the ability to write multiple files to the local disk and can trigger outbound communications (emails) if the default safety mode is overridden.
Sanitization: The playbook does not describe any sanitization or filtering logic for the data retrieved by the scrapers before it is used for strategy synthesis and drafting.
[EXTERNAL_DOWNLOADS]: The skill utilizes several external data gathering capabilities.
Evidence: Incorporates tools such as apollo-lead-finder, review-scraper, and web-archive-scraper to fetch data from various internet sources as part of its intelligence gathering phase.
[COMMAND_EXECUTION]: The skill provides a mechanism to disable its primary safety constraint.
Evidence: The pitch_packet_mode: false configuration override explicitly enables live campaign execution, which includes sending real emails and spending paid API credits (SKILL.md, Configuration Overrides).

client-packet-engine