The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill makes extensive use of npx @vibeprospecting/vpai@latest to perform account discovery and enrichment. This command downloads and executes the latest version of a third-party CLI tool from the npm registry at runtime without version pinning or integrity verification.
[PERSISTENCE]: In references/06-campaign-management.md, the skill instructs the user to add a monitoring script to the system's crontab (0 8 * * * ...). This establishes a persistence mechanism that ensures code is executed automatically and periodically on the host machine.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of playwright and the subsequent download of browser binaries (playwright install chromium). While Playwright is a well-known tool, the runtime installation of external binaries increases the attack surface.
[COMMAND_EXECUTION]: The skill orchestrates multiple phases using shell commands and Python subprocesses. It dynamically generates and executes scripts to handle data processing and API interactions with LinkedIn, which can be risky if inputs are not strictly controlled.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing company data from external sources (Vibe Prospecting) to generate ad copy and headlines.
Ingestion points: Data enters through abm-accounts-raw.csv and JSON enrichment files in Phases 1 and 2.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded within the company names or firmographic data.
Capability inventory: The agent has the capability to execute shell commands, write files, and configure system tasks (cron).
Sanitization: The skill lacks evidence of sanitization for company names or domains before they are interpolated into scripts, shell commands, or ad templates.

abm-campaign