The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it is designed to ingest and process untrusted external data (such as competitor pricing pages via agent-browser) and incorporates this data into its workflow.
Ingestion points: Competitive pricing pages (via agent-browser research mode) and local strategic context files (brand-context.md, paw-mkt-product-context.md, etc.).
Boundary markers: No explicit instruction wrappers or delimiters are defined to isolate untrusted data from the system prompt.
Capability inventory: The skill possesses network access (via agent-browser), the ability to execute shell scripts (e.g., tool-discovery.sh), and file-writing capabilities (deliverable generation).
Sanitization: The instructions do not specify any sanitization, validation, or escaping of external content before processing.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install the agent-browser tool from Vercel Labs' official GitHub repository (github.com/vercel-labs/agent-browser). This is a trusted source and is documented as a necessary dependency for the skill's research capabilities.
[COMMAND_EXECUTION]: The skill utilizes shell commands and scripts to perform environment checks and execute research tasks.
Tool discovery: Runs ./skills/paw-mkt-setup/assets/scripts/tool-discovery.sh to verify the presence of required runtimes and tools.
Profile discovery: Executes ./skills/paw-mkt-setup/assets/scripts/chrome-profiles.sh to identify browser profile paths for authenticated sessions.
Browser automation: Uses agent-browser to navigate to external URLs, take screenshots, and extract text for competitive analysis.

paw-mkt-pricing