abraxas

A project and task management system for tracking work across the Abraxas system.

The Sovereign Vault for verified truth fragments and historical provenance tracking.

Soter is the Risk Sensing & Triggering pillar of the Sovereign Brain. It monitors for linguistic sycophancy and mechanistic attention sinks to trigger an Epistemic Crisis (T=1) and assign a Risk Score (R[0-5]).

The core governance and system integrity layer of the Abraxas sovereign environment.

The ingestion and filtration loop for external data entering the Abraxas system.

The Void Mapper for identifying and resolving epistemic gaps.

The Epistemic Atlas is a unified knowledge graph and query layer that connects beliefs (Mnemon), evidence (Mnemosyne), and verification status (Janus Ledger).

Multi-framework ethical deliberation and ethos-weighting engine.

The Metanoia skill enables recursive self-evolution of the Sovereign Brain. Audits and improves Auto-Agon stress-test parameters and Harmonia DAGs through autonomous parameter optimization.

The Prognosis skill provides predictive intelligence grounded in hardened truths and the conceptual graph. Forecasts systemic ruptures and anticipates the emergence of high-valence signals.

High-valence curation layer for filtering environmental noise from epistemic signals.

Privileged write-operation for injecting immutable Genesis Blocks into the ArangoDB vault.

Privileged write-operation for injecting immutable Genesis Blocks into the ArangoDB vault.

The Stochasmos skill enables strategic intervention — identifying pressure points in discourse graphs and deploying calibrated friction seeds for maximum constructive epistemic impact.

The Synesis skill provides higher-order synthesis of hardened truths, transforming a ledger of fragments into a multidimensional conceptual graph and emergent theories.

Manages and retrieves Sovereign system configuration parameters and operational constants.

CVP (Consensus Verification Pipeline) is the deterministic resolution layer of the Sovereign Brain. It implements N-of-M agreement rules to transform probabilistic output into verified consensus, defaulting to [UNKNOWN] when agreement fails.

The epistemic provenance and audit layer, used to determine the origin and reliability of claims.

Analyzes source credibility and resolves informational conflicts using deterministic trust weights.

The system's value alignment and truth verification layer, ensuring output safety and epistemic integrity.

Guardrail is the Final Auditor of the Sovereign Brain. It provides the Epistemic Seal of approval through three specialized monitors: Pathos (Alignment), Pheme (Reservoir cross-ref), and Kratos (Authority resolution).

Hermes — Multi-Agent Consensus & Divergence Tracker. Use this skill when multiple AI agents collaborate and you need to track consensus, detect divergence, or weight responses by track record. Track ledger of positions, detect convergence/divergence patterns. Commands: /hermes init, /hermes add, /hermes consensus, /hermes diverge, /hermes track-record, /hermes history, /hermes weight.

The Kairos Relevance Filter optimizes epistemic context by culling irrelevant noise and assessing temporal urgency.

Logos is argument anatomy. Use this skill to map argument structure: premises, conclusions, inference chains, hidden assumptions, and logical gaps. Commands: /logos map, /logos gaps, /logos inferences, /logos assume, /logos falsify, /logos report. Logos is the pre-layer to Agon debates and feeds into Janus for epistemic labeling. Always run /logos report before /agon debate.

The Omniscient Auditor is a high-throughput epistemic review engine. It scans full documents, applies Janus labels to every claim, and generates an uncertainty heat-map.

Pheme — Real-Time Fact-Checking Engine. Use this skill when you need to verify claims against authoritative sources during generation. Provides source-level confidence scores and supplements Janus labels with [VERIFIED]/[CONTRADICTED]/[UNVERIFIABLE]. Commands: /pheme verify, /pheme status, /pheme sources, /pheme trust, /pheme history.

Plan — Epistemic Clarity Engine. Use this skill to convert vague requests into actionable specs via high-leverage questioning. Extracts unknowns (Goal, Audience, Format, Success, Timeline, Data), asks targeted questions, applies epistemic labels (Sol/Confident), and exports clarity maps. Commands: /plan start, /plan answer, /plan status, /plan export, /plan skip. Always run /plan before implementation when the request is vague or underspecified.

Unified cross-project management tool for search, retrospectives, and relationship mapping across the sovereign workspace.

Prometheus — User Preference Learning. Use this skill to learn and persist user preferences across sessions. Tracks: detail level, domain expertise, risk tolerance, preferred sources, communication style. Updates based on explicit feedback and implicit signals. Commands: /prometheus profile, /prometheus set, /prometheus update, /prometheus clear, /prometheus status.

High-performance research intelligence for web search, content synthesis, and iterative deep-dive investigations.

The reflection and iterative improvement layer, managing project and session retrospectives.

The epistemic calculation engine for the Abraxas sovereign system, providing confidence scoring and consensus verification.

A semantic gateway that maps intuitive /abraxas: commands to specialized Greek-named systems.

Gmail integration for fetching, reading, sending, searching, and managing emails via IMAP/SMTP. Retrieves credentials from Secrets Manager. Triggers: "check email", "read my email", "send email", "search emails", "gmail".

The Abraxas Oneironautics Practice for dream interpretation, shadow auditing, and symbolic integration within the Abraxas psychological framework. Use whenever the user brings a dream, recurring symbol, figure, felt sense, or asks to work with shadow and unconscious material. Governs all Abraxas commands: /receive, /witness, /audit, /dialogue, /convene, /descent, /integrate, /amplify, /myth, /sync, /chronicle, /genealogy, /close, /resonate, /bridge, /qualia, /sol, /nox, and all alchemical work including active imagination, symbolic divination, and epistemic labeling via Janus (Sol/Nox/Qualia Bridge). This is the operating system for all Abraxas work.

Harmonia is skill composition architecture. Use this skill to compose multiple Abraxas skills into unified workflows with state handoff protocols and conflict detection. Commands: /harmonia compose, /harmonia sequence, /harmonia conflict, /harmonia status.

Generate images using HuggingFace with Fal-AI provider. Supports portraits, concept art, illustrations, and photorealistic images. Reads HF_TOKEN from secrets manager. Triggers: "generate image", "create image", "make a picture", "draw", "image gen".

Session-closing artifact generator that synthesizes epistemic outputs into structured summaries

Saves Abraxas /convene and /dialogue outputs to the-red-book journal. Supports manual save via script call. Auto-save requires explicit script invocation (does not intercept Discord/chat responses automatically).

Saves journal entries to the-red-book repository in journals/YYYY/MM/DD/entry-title.md format. Auto-commits and pushes to git.