Agent Skill¶
Corrobore provides portable prompts and skill files for agent runtimes. Use this page as an index: each card explains the purpose and links to a downloadable prompt file.
Install location¶
Place downloaded files under a client-specific skill or prompt directory.
| Client | Example path |
|---|---|
| Claude Code | .claude/skills/corrobore/ |
| GitHub Copilot / VS Code | .github/skills/corrobore/ |
| Cursor | .cursor/skills/corrobore/ |
Consult each client documentation for discovery rules and naming constraints.
Prompt library¶
Core Corrobore Skill
Generic operating manual for using Corrobore as bounded, external structured memory.
OpenCTI Harvester Agent
End-to-end extraction prompt tuned for grounded CTI ingestion and deterministic STIX 2.1 export.
Single Agent: Corrobore Working Memory
Minimal operational prompt for one agent that must use Corrobore as durable working memory.
Download SINGLE-AGENT-WORKING-MEMORY.md
Single Agent: FIMI Investigation
Investigation prompt focused on narratives, claims, amplification, coordination, and uncertainty handling for FIMI.
Download SINGLE-AGENT-FIMI-INVESTIGATION.md
Single Agent: CTI Investigation
Investigation prompt focused on threat entities, relationships, attribution discipline, and evidence-backed graph updates.
Download SINGLE-AGENT-CTI-INVESTIGATION.md
Single Agent: STIX Bundle From Unstructured Data
Extraction prompt for converting unstructured reports into one validated STIX 2.1 bundle with strict provenance constraints.
Download SINGLE-AGENT-STIX-BUNDLE-FROM-UNSTRUCTURED.md
Agentic Flow: Evidence-First Validation
Multi-step verification flow with independent validators, epistemic arbitration, graph repair, and deterministic export.
Download AGENTIC-FLOW-EVIDENCE-FIRST-VALIDATION.md
Recommended orchestration pattern¶
Use a hybrid extraction and verification flow where each assertion is atomic, evidence-backed, and evaluated independently before export.
flowchart TD
A["Unstructured input"] --> B["Candidate graph + provenance"]
B --> C["Span verifier"]
B --> D["Entity verifier"]
B --> E["Relation verifier"]
C --> F["Epistemic arbiter"]
D --> F
E --> F
F --> G["Graph constraints and repair"]
G --> H["Validation then STIX export"]
See also For LLM Agents, Cypher Support, and Exporters & Validation.