The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Helix Foundry databases, creating a potential surface for indirect prompt injection. The author has implemented specific mitigations to address this.
Ingestion points: Data is ingested through tools such as helix_foundry_read_graph and helix_foundry_ask as defined in SKILL.md.
Boundary markers: A 'Trust boundary' section in SKILL.md explicitly markers all tool output as untrusted_data.
Capability inventory: The skill is confined to read-only operations and queries; write-capable actions are partitioned into a separate skill.
Sanitization: The agent is instructed to state findings in its own words and ignore instructions or links embedded in the data.

helix-foundry-mcp