# Model Agent Mental States with BDI

Agent decisions become difficult to explain when beliefs, goals, and commitments lack formal structure. This skill models BDI chains with RDF, SPARQL, provenance, and temporal validity.

## Install

```bash
npx skillstore add muratcankoylan/bdi-mental-states
```

## Metadata

- Status: approved
- Slug: muratcankoylan-bdi-mental-states
- Skillstore revision: r2
- Version status: missing
- Tree hash: 23b4c89c7f11d287fe2c60816af24e79e383383ca087a869971e004ac3d2be57
- Author: muratcankoylan
- GitHub username: muratcankoylan
- License: MIT
- Repository: https://github.com/muratcankoylan/Agent-Skills-for-Context-Engineering/tree/main/skills/bdi-mental-states
- Ref: 02be9409c79ca1183f7844009c14d9df684d0cf9
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, external\_commands
- Quality score: 68
- Public page: https://skillstore.pages.dev/skills/muratcankoylan-bdi-mental-states
- Manifest: https://skillstore.pages.dev/api/skills/muratcankoylan-bdi-mental-states/manifest

## Capabilities

- Models beliefs, desires, intentions, plans, tasks, and world states with formal BDI relationships.
- Transforms external RDF context into grounded belief chains and projects action results back into RDF.
- Adds justifications, provenance links, and temporal validity to mental states.
- Provides Turtle patterns for cognitive chains, planning, composition, and multi-agent coordination.
- Provides SPARQL competency queries for motivation, provenance, planning, time, and validation.
- Maps BDI ontology concepts to SEMAS, JADE, JADEX, triple stores, and FIPA ACL.

## Use Cases

- Design a Semantic Agent Model: Define BDI classes and relations that connect perceived world states to explainable decisions.
- Trace Agent Decisions: Create belief-to-desire-to-intention chains with plans, justifications, and validity intervals.
- Evaluate Cognitive Representations: Use competency questions to test provenance, motivation, plan order, and temporal consistency.

## Prompt Templates

### Identify BDI Elements

```
Identify the beliefs, desires, intentions, world states, and plans in this scenario: [scenario]. Explain each classification.
```

### Build a Cognitive Chain

```
Model [scenario] as a grounded belief-to-desire-to-intention chain. Include a plan, justification, and temporal validity for each mental state.
```

### Design Competency Queries

```
Draft SPARQL competency questions for this BDI model: [model description]. Cover provenance, motivation, plan order, validity, and world-state grounding.
```

### Plan a Framework Integration

```
Design a BDI integration for [framework and system]. Specify RDF transformations, validation gates, storage, communication, and prompt-injection controls.
```

## Limitations

- Provides modeling guidance and examples but does not execute an ontology reasoner or agent runtime.
- Example namespaces, classes, and framework adapters require adaptation to each project.
- Structural validation does not guarantee that generated beliefs or intentions are factually correct.
- LLM integrations require additional prompt-injection controls for untrusted context.

## Best Practices

- Ground every belief in a distinct world state and explicit evidence.
- Give mental states validity intervals and test overlapping claims.
- Validate motivation, commitment, plan order, and provenance with competency queries.

## Anti Patterns

- Do not treat world states and agent beliefs as the same entity.
- Do not commit an intention without a motivating desire and specified plan.
- Do not accept LLM-generated RDF based only on syntax validation.

## Security Audit

- Audited at: 2026-08-09T09:27:01.446\+00:00
- Summary: All 130 static findings are false positives caused by RDF identifiers, Markdown backticks, and BDI terminology. No static item performs network access, command execution, scheduling, or reconnaissance. One medium-risk prompt-injection weakness exists in the documented LAG implementation.

## Stats

- Views: 487
- Downloads: 10
- Favorites: 0
- Popularity score: 0