The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and analyzes content from external company websites to perform a 'Quality Checklist'. Maliciously crafted content on these websites could potentially influence the agent's decision-making process during the QA review.
Ingestion points: Uses WebFetch to retrieve content from company domains (company.domainName.primaryLinkUrl) and demo playground URLs.
Boundary markers: The instructions do not define specific delimiters or instructions for the agent to ignore potentially malicious content embedded in the fetched HTML.
Capability inventory: The agent has the ability to execute curl commands for API interaction and run direct SQL queries via a Supabase MCP plugin.
Sanitization: There is no mention of sanitizing or filtering the fetched web content before it is used to evaluate criteria like 'Greeting quality' or 'No hallucinations'.
[COMMAND_EXECUTION]: The skill utilizes curl to perform GraphQL mutations on crm.psquared.dev and to call management APIs at app.psquared.dev. These commands are used to transition demo statuses and trigger widget style updates.
[DYNAMIC_EXECUTION]: The skill uses the mcp__plugin_supabase_supabase__execute_sql tool to perform direct UPDATE operations on the demo_pages table. While this is identified as a necessary workaround for missing API tools, executing raw SQL based on values derived from the QA process (which ingest untrusted data) is a sensitive operation.
[DATA_EXPOSURE]: The skill accesses the .env file to retrieve sensitive tokens (PSQUARED_CRM_TOKEN, NUXT_MCP_DEMO_TOKEN, OPENBRAND_API_KEY). While using an environment file is a standard way to manage secrets, the skill explicitly reads these credentials to authenticate its operations against vendor and third-party APIs.

review-demos