The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses sensitive authentication sessions stored at {project-root}/.pawbytes/creative-suites/.auth/ to enable authenticated browser research on platforms like Instagram and TikTok. It also reads configuration files from {project-root}/.pawbytes/config/ which are expected to contain environment or user-specific settings for the pawbytes ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection due to its core function of crawling and summarizing untrusted external web content.
Ingestion points: The skill ingests data from external sources via crawling_exa, web_search_exa, and social media feed scrolling using the agent-browser CLI.
Boundary markers: The instructions do not define specific delimiters or instructions to the agent to disregard potentially malicious commands embedded within the crawled competitor websites or social media posts.
Capability inventory: The agent can write research reports and daily log entries to the local file system based on the processed external data.
Sanitization: There is no mention of sanitization, filtering, or validation of the content retrieved from external URLs before it is synthesized into production recommendations.
[COMMAND_EXECUTION]: The skill involves the execution of external tools and command-line interfaces, specifically the Exa MCP tools and the agent-browser CLI, to perform its research tasks.

paw-cra-content-research