The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill mandates fetching and processing content from external documentation sites to drive its behavior and code generation.
Ingestion points: SKILL.md and references/search-first.md instruct the agent to fetch content from https://docs.stigg.io/llms.txt and various sub-paths under docs.stigg.io to retrieve current API schemas and SDK patterns.
Boundary markers: Absent. The instructions command the agent to treat the fetched content as authoritative for its output.
Capability inventory: The agent uses this data to generate backend/frontend integration code and to formulate queries for the Stigg MCP server.
Sanitization: Not explicitly defined; the skill relies on the agent's ability to safely parse documentation content.
[PROMPT_INJECTION]: The skill employs strong instructional directives intended to override the agent's default reasoning and reliance on static training data.
Evidence: SKILL.md labels the "Search-First Rule" as "non-negotiable" and uses phrases like "Red flag — STOP and search" and "delete it and start over" to strictly enforce a specific operational flow. This is used as a quality control mechanism rather than a bypass of safety filters.
[REMOTE_CODE_EXECUTION]: The skill documents the installation of the Stigg CLI and the connection to the Stigg MCP server.
Evidence: SKILL.md and references/decision-tree.md refer to brew install stiggio/tools/stigg and the use of the official MCP server at mcp.stigg.io. These are verified vendor resources belonging to the skill author.

stigg