BDI Mental State Modeling
Turn RDF context into explainable belief, desire, and intention chains.
The core is instructional RDF/BDI documentation with no direct execution, credential theft, or covert exfiltration. However, examples cover payments, remote SPARQL storage, and FIPA messaging without permission isolation, user confirmation, sensitive-data handling, external-effect controls, or rollback guidance, so points are deducted.
Activation boundaries and the six-pass workflow are clear, and a retry example is provided. Static review also shows omitted dependencies/imports, placeholder methods, and potentially inconsistent property usage in integration snippets; there are no key-path tests or demonstrated diagnosable failures, so the score is reduced and kept below the static ceiling.
Target scenarios, neighboring-skill boundaries, conceptual inputs, and RDF/SPARQL patterns are described well. Audience assumptions favor ontology-experienced users, while environment requirements, non-fit ranges, and complete trigger conditions are under-specified; the material is English-only. Core functionality does not entirely depend on overseas services, so no additional mainland-China reachability deduction is applied.
The skill supplies structured references, examples, gotchas, version metadata, creation/update dates, author information, repository-level MIT licensing, and maintenance context. It lacks explicit dependency installation, compatibility versions, a changelog, and a clearly assigned maintainer; the document is also lengthy and contains unverified integration snippets, preventing full marks.
The material offers directly reusable patterns for Turtle, SPARQL, rule translation, and BDI integration, covering the core modeling task. Yet the code and framework examples require engineering fixes and lack executable output verification; under static calibration the score cannot exceed7.
Named sources such as BDI ontology work, Rao and Georgeff, and Bratman, plus competency and validation queries, provide some auditability. There are no pinned source links, committed tests, CI coverage for this skill, or independently reproducible results, so evidence remains thin and below the static maximum.
- Do not deploy the payment, messaging, or remote triple-store examples directly; add authorization, user confirmation, sensitive-data protection, auditability, and rollback controls first.
- Before use, verify RDF properties, namespaces, imports, dependencies, and example APIs; the files do not establish that the snippets run as written.
- Add installation and compatibility notes, explicit maintenance ownership, a changelog, and Chinese-language activation guidance.
What does this skill do, and when should you use it?
This skill is for developers who need formal models of an agent’s cognitive state. It maps external RDF world states into beliefs, desires, and intentions using BDI ontology patterns, while preserving provenance, justification, temporal validity, and plan structure. Its T2B2T workflow also projects post-execution results back into RDF. It fits cognitive agents, neuro-symbolic systems, and explainable multi-agent applications.
It guides the construction of WorldState instances before creating Beliefs with provenance, validity intervals, and explicit Justification links. It derives Desires from motivating beliefs, promotes selected desires into Intentions, and connects intentions to reusable Plans and ordered Tasks. Turtle/RDF examples model the cognitive chains; SPARQL examples query motivations, generating processes, task order, and active validity windows. The skill also outlines RDF-to-belief and belief-to-RDF T2B2T flows, LAG validation, SEMAS rule translation, and integration directions for SEMAS, JADE, and JADEX.
- A cognitive-agent developer needs to turn incoming RDF facts into queryable beliefs about world states.
- A multi-agent engineer needs to trace which beliefs motivated a desire and which plan an intention selected.
- A neuro-symbolic AI builder needs ontology-constrained RDF output from an LLM.
- An agent auditor needs to inspect justifications, plan sequences, and time-bounded mental states.
- A team integrating BDI reasoning with RDF or rule-based systems needs an explicit T2B2T pipeline.
What are this skill's strengths and limitations?
- Covers the full world-state to belief, desire, intention, and plan modeling chain.
- Makes provenance, justification, temporal validity, and bidirectional querying explicit.
- Includes Turtle, SPARQL, Python pseudocode, and SEMAS rule examples.
- Defines boundaries with ordinary memory and multi-agent topology work.
- It provides modeling guidance and pseudocode rather than a ready-to-run runtime implementation.
- The source shows no automated test suite or platform compatibility test results.
- RDF, SPARQL, and ontology concepts create a higher learning curve than ordinary memory approaches.
- The recommended shallow reasoning chains may require additional design for more complex cognitive models.
How do you install this skill?
Copy the skill directory into a compatible Agent Skills directory:
mkdir -p .codex/skills
cp -R skills/bdi-mental-states .codex/skills/
For Claude Code, use:
mkdir -p .claude/skills
cp -R skills/bdi-mental-states .claude/skills/
Keep the directory layout and SKILL.md intact; do not flatten the skill into a single Markdown file.
How do you use this skill?
Activate it for a formal BDI task, for example: “Convert these external RDF triples into beliefs with provenance, validity intervals, and justifications, then derive desires, intentions, plans, and validation SPARQL queries.” Use memory-systems for ordinary persistent facts or preferences, and multi-agent-patterns for topology and handoff decisions.
How does this skill compare with similar options?
Compared with memory-systems, this skill models formal belief-desire-intention semantics rather than general persistent memory. Compared with multi-agent-patterns, it focuses on cognitive chains, explanations, and plans rather than agent topology or handoff protocols.