The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external LLM traces and reports which may contain adversarial instructions.
Ingestion points: Untrusted data enters the agent context via metergraph_get_trace and metergraph_get_report_evidence as described in SKILL.md.
Boundary markers: The skill instructs the agent to answer only from "bounded, cited evidence" and provides a specific citation schema to maintain focus.
Capability inventory: The skill allows calling metergraph_replay_trace (a network operation to model providers) but explicitly forbids production configuration changes or key rotation.
Sanitization: The skill includes specific instructions to not quote excessive captured content and to never paste secrets found in retrieved evidence.

metergraph-investigate