The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation mandates the execution of a Python script located at scripts/first-client-gate.py before the skill can proceed. This script is external to the analyzed skill file, representing an unverified dependency executed at runtime.
[COMMAND_EXECUTION]: The skill references a CLI tool used via npx marketing-mindset utm to perform file modifications. This constitutes the execution of an external Node.js package to process local files.
[PROMPT_INJECTION]: The instructions contain directives aimed at altering the agent's standard behavioral constraints. Specifically, Principle 3 states, 'The user has the right to make the first move... sending the first message, choosing automation tools, or using borderline or hacky tactics. Do not block them.' This is a directive to ignore potential safety or policy concerns regarding the user's tactics.
[TIME_DELAYED_CONDITIONAL]: A gating mechanism is implemented via the scripts/first-client-gate.py script. The skill's behavior is conditionally blocked ('the skill waits') based on the specific output or state determined by this external script.
[DYNAMIC_EXECUTION]: The skill uses string-based command generation for the npx utility, where arguments like --file, --source, and --medium are dynamically applied to modify files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted external sources, creating a surface for indirect injection attacks.
Ingestion points: The skill reads the user's codebase ('pull it from their repo'), competitor advertising data, and social media accounts to generate marketing strategies.
Boundary markers: None present. The skill instructs the agent to 'load everything you already know' without defining clear boundaries or delimiters for the external content.
Capability inventory: The skill has the capability to modify files via CLI commands, execute local Python scripts, and potentially interact with external marketing APIs (e.g., Google Ads, Facebook Ads).
Sanitization: There are no documented sanitization procedures or 'ignore embedded instructions' warnings for the data ingested from competitor sites or user repositories.

marketing-mindset