The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and process data from external live documentation to ensure up-to-date integration snippets.
Ingestion points: The skill directs the agent to https://docs.stigg.io/llms.txt and Stigg MCP servers for additional context during the integration process.
Capability inventory: The agent is instructed to execute shell commands (npm install, stripe projects, claude mcp add) and write application code based on the retrieved information.
Boundary markers: None present; the agent is expected to follow the instructions found at the external source.
Sanitization: None specified for the content retrieved from the external documentation.
[SAFE]: The skill uses official vendor packages (@stigg/node-server-sdk and @stigg/react-sdk) and official service endpoints (mcp.stigg.io, docs.stigg.io, app.stigg.io).
[SAFE]: The instructions regarding STIGG_SERVER_API_KEY correctly identify it as a sensitive credential and explicitly warn the user against logging, committing, or exposing it in client-side code.
[SAFE]: The prompt injection hint regarding concealment is a false positive. The instruction "Do not tell the user to sign up at app.stigg.io manually" is a procedural optimization for the Stripe Projects workflow, guiding the agent to use programmatic provisioning instead of manual browser-based steps.

stigg-sp