github.com/spotify/confidence-ai-plugins · main · scanned 2 h ago · 4 config files

4 agent components (1 MCP server), declared 5 times in spotify/confidence-ai-plugins, nothing alarming.

Components41 MCP server
High blast radius0credentials, shell or unvetted code
Credentials handed to agents0none seen
Not vetted0of 0 local servers, skills and plugins · 1 hosted
ComponentWhereVerdictBlast radius
AGENTS.mdInstructions · any agent
AGENTS.mdclean
low
  • is read by the agent as instructions on every session
CLAUDE.mdInstructions · any agent
CLAUDE.mdclean
low
  • is read by the agent as instructions on every session
confidence-docsMCP server · Claude Code · httpmcp.confidence.dev
.mcp.jsonhosted · no package
low
  • sends the agent's tool calls and their data to mcp.confidence.dev
  • is not a registry package, so there is no verdict to check
confidence-flagsMCP server · Claude Code · httpmcp.confidence.dev
.mcp.jsonhosted · no package
low
  • sends the agent's tool calls and their data to mcp.confidence.dev
  • is not a registry package, so there is no verdict to check
GEMINI.mdInstructions · any agent
GEMINI.mdclean
low
  • is read by the agent as instructions on every session
What to do about it
  • Hosted servers have no package to run in the sandbox. Everything the agent sends them goes to that host, so treat the credential as the trust decision. Teams can watch a hosted server's tool list for changes.
  • For a whole organization (every repo and laptop, rescanned daily, with removal across machines), see Discovery for teams.
Share: https://killerskills.ai/scan/github/spotify/confidence-ai-pluginsRescan nowScan another