The Agent Skills Directory

[COMMAND_EXECUTION]: Executes multiple curl commands to interact with vendor-specific infrastructure (psquared.dev) and CRM systems for data retrieval and agent management.
[COMMAND_EXECUTION]: Provides instructions for running a local Python script (scripts/refurbish-all.py) to handle batch processing of agent updates.
[CREDENTIALS_UNSAFE]: Accesses sensitive Bearer tokens from environment variables (NUXT_MCP_DEMO_TOKEN, PSQUARED_CRM_TOKEN, EMAIL_DRAFT_ONLY_BEARER) to authenticate requests to internal services and the CRM.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted external URLs to populate agent knowledge bases, creating a vulnerability surface for indirect prompt injection.
Ingestion points: External web pages identified through domain patterns or WebFetch (found in SKILL.md).
Boundary markers: None explicitly specified to prevent the agent from following instructions embedded in the scraped content.
Capability inventory: Modifies knowledge buckets, updates agent configuration, and republishes agents via MCP tools (found in SKILL.md).
Sanitization: The skill describes removing navigation elements and footers but lacks specific sanitization logic to neutralize malicious instructions within the scraped text.

refurbish-demos