The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local shell and batch scripts located in a directory outside the skill's own folder, specifically ./skills/paw-mkt-setup/assets/scripts/tool-discovery.sh and ./skills/paw-mkt-setup/assets/scripts/chrome-profiles.sh (found in references/shared-patterns.md).
[DATA_EXFILTRATION]: The skill provides instructions for accessing and managing sensitive browser session data. It directs the agent to discover local Chrome profile paths (e.g., ~/.linkedin-profile) and to save authentication states to files like ./my-auth.json, which the skill explicitly notes contain 'sensitive session tokens in plaintext' (found in references/shared-patterns.md). This creates a risk of credential exposure if the agent is compromised.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its web scraping capabilities (found in references/research-mode.md and references/shared-patterns.md).
Ingestion points: The agent fetches and processes text from external websites including Google News, LinkedIn, HARO, and competitor sites via agent-browser.
Boundary markers: There are no instructions to use delimiters or ignore embedded instructions when processing fetched web content.
Capability inventory: The agent has the ability to execute shell scripts, write files to the brand workspace, and perform further browser-based interactions.
Sanitization: There is no evidence of sanitization or validation of the content retrieved from external URLs before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill recommends installing the agent-browser tool from Vercel Labs' GitHub repository using npx (found in references/shared-patterns.md).

paw-mkt-pr