swarm-ai-research

Audit every enabled skill's upstream file dependencies for staleness — flags chained skills about to consume yesterday's article or a long-dead topic file

Audit .github/workflows and composite actions with zizmor + actionlint, classify findings against the prior audit, auto-fix Critical/High regressions, and open a PR only when something actually changed.

Weekly API cost report — computes dollar costs from token usage, flags anomalies, forecasts burn, and prescribes concrete optimizations

Weekly partial-correlation of compute economics against a Hyperliquid macro basket — DePIN-token proxy track runs every week (n>180d), sweep-P&L track defers until n≥30 joined days

Score frontier LLMs (Claude, GPT, Gemini, DeepSeek, Grok) on a private compute-markets task corpus, track score deltas across releases, flag public-vs-private divergence

Save a note as one or more atomic notes under memory/notes/ (and optionally Supernotes). Splits bundled inputs into separate atomic files.

Regenerate notegraph.json + docs/notegraph.md + docs/notegraph.html from memory/ and docs/, PR + notify only when the graph changes

Review recent activity, consolidate memory into atomic notes + MOCs, and prune stale entries

SWARM: System-Wide Assessment of Risk in Multi-agent systems. Simulate multi-agent dynamics, test governance, study emergent risks.

Query the SWARM knowledge graph structurally — find pages, list backlinks, follow link paths, and surface related/semantic neighbors across docs, scenarios, slash commands, agents, roles, and code references. Prefer this over grep when you need *connected* answers ("what links to X", "how does X relate to Y", "what's similar to X").

Scaffold research papers from SWARM run data with auto-populated tables

Run parameter grid sweeps across SWARM scenarios and generate summary statistics

Generate publication-quality visualizations from SWARM simulation data

Execute a SWARM simulation scenario and export standardized artifacts

Perform rigorous statistical analysis on SWARM experiment data with multiple-comparison corrections

Create or update a claim card in the vault. In create mode, guides through evidence collection, confidence assessment, and boundary conditions. In update mode, adds new evidence or changes status. Triggers on "/claim", "/claim create [proposition]", "/claim update [claim-id]", "/claim status", "add a claim", "update claim evidence".

Find connections between claims and update topic maps. Requires semantic judgment to identify genuine evidence relationships. Use after /extract creates claims, when exploring connections, or when a topic map needs synthesis. Triggers on "/cross-link", "/cross-link [claim]", "find connections", "update topic maps", "connect these claims".

Extract structured claims from SWARM experiment runs, sweep results, and red-team reports. Comprehensive extraction is the default — every governance finding, adversarial pattern, and parameter sensitivity result gets extracted. Zero extraction from a domain-relevant run is a BUG. Triggers on "/extract", "/extract [file]", "extract findings", "mine this run", "process this experiment".

Explore and analyze the SWARM knowledge graph — detect orphan claims, find dense connection clusters, trace evidence chains, identify synthesis opportunities. Triggers on "/graph", "graph analysis", "find orphans", "what connects to X".

Research a topic relevant to SWARM AI safety and grow the vault with findings. Queries external sources, synthesizes into claim candidates, follows the extraction pipeline. Triggers on "/learn", "/learn [topic]", "research this", "find papers on".

Surface the next highest-priority action from the SWARM synthesis pipeline queue and maintenance conditions. Recommends what to work on based on queue state, pending observations, stale claims, and research threads. Triggers on "/next", "what should I work on", "next action", "what's pending".

Run the full SWARM synthesis pipeline on one or more experiment runs — seed, extract, cross-link, update, validate in sequence. Triggers on "/pipeline", "/pipeline [run-folder]", "full pipeline", "process everything".

Capture methodology friction, process corrections, and operational learnings for the SWARM vault. Rule Zero — methodology notes are the spec; when something goes wrong, write it down so the system learns. Triggers on "/remember", "capture friction", "log this correction", "the system did X wrong".

Challenge SWARM vault assumptions against accumulated observations and tensions. Triages friction signals, detects patterns, proposes methodology updates. The scientific method applied to the research OS itself. Triggers on "/rethink", "review observations", "challenge assumptions", "system drift".