The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external materials (such as pitch decks, customer research, and kickoff transcripts) provided by a client and stored in a local materials/ directory. This creates a potential surface for indirect prompt injection where malicious instructions inside these documents could attempt to influence the agent's behavior or bias the generated strategy.
Ingestion points: METHODOLOGY.md (Step 1.2) specifies reading all client-provided materials.
Boundary markers: Not explicitly defined for ingested content.
Capability inventory: File reads and writes within the ~/marketing-plans/ directory; network operations via authorized MCPs (e.g., Stripe, GA4, Ahrefs).
Sanitization: Content is not sanitized, though the file paths are protected by strict slugification rules.
[COMMAND_EXECUTION]: The documentation in OPS-STACK.md identifies several external CLI tools (e.g., agent-browser, defuddle, obsidian-cli) as part of the recommended marketing operations stack. The skill provides instructions for the user to install these tools (e.g., npm install -g agent-browser) but does not contain instructions for the agent to execute arbitrary or unauthorized shell commands.
[EXTERNAL_DOWNLOADS]: The skill references several external resources, primarily documentation and linked skills hosted on GitHub within the Quegenx/agent-skills and coreyhaines31/marketingskills repositories. These references are used for methodology and context and are considered safe educational links.

marketing-plan