// Append notable events to state/memory.jsonl. Use after meaningful interactions, discoveries, decisions, or state changes. Forgetting runs via maintenance script.
Use when an experiment run has completed and the user wants rigorous interpretation, claim-driven charts, bilingual reports, or cross-hypothesis synthesis from simulation data. Also use when multiple charts or PNG/JPG assets must be assembled into one labeled composite figure. Requires high-quality narrative and evidence traceability, not only harness gate PASS.
Create distinctive, production-grade frontend interfaces with high design quality. Use when building or polishing analysis HTML reports under agentsociety-analysis — read via support/frontend-design/ inside that skill, not as a separate pipeline skill.
Multi-modal interactive data presentation for agentsociety-analysis — EDA bundle (PyGWalker, Plotly, sortable tables, eda_hub), plotly/altair claim charts, HTML tab surfaces. Use in explore, refine, and produce stages.
Composable HTML block patterns for agentsociety-analysis reports — KPI strips, figure cards, Mermaid, EDA tabs, optional interactive chart iframes. Use during Stage 5 produce when authoring report_zh.html / report_en.html.
Publication-quality chart patterns for agentsociety-analysis Stage 4 refine — Okabe-Ito palettes, seaborn CI bands, small multiples, error bars, grayscale-safe encoding. Use when writing run-code chart scripts or reviewing chart QA failures.
Use when starting or resuming an AgentSociety research workspace, deciding which research skill to invoke next, checking current pipeline state, or sizing a simulation before configuration and module creation.
| name | memory |
| description | Append notable events to state/memory.jsonl. Use after meaningful interactions, discoveries, decisions, or state changes. Forgetting runs via maintenance script. |
| script | scripts/memory_maintenance.py |
Store at most one durable memory from the current tick in state/memory.jsonl.
This skill decides what is worth remembering. It does not retrieve memories for reasoning, and it does not reimplement forgetting. Forgetting is handled by scripts/memory_maintenance.py.
Use after observation, cognition, planning, action, or social interaction when something may matter later.
state/observation.txtAppend one JSON object per line:
{"tick":42,"time":"2024-01-15T10:30:00","type":"event","summary":"Met Alice at the park; she mentioned a library job.","tags":["social","alice","job"],"importance":"medium"}
Required fields:
| Field | Notes |
|---|---|
tick | Current tick, if available |
time | ISO timestamp, if available |
type | need, emotion, cognition, intention, plan, plan_execution, react, event, observation, social, decision, discovery, or plan_outcome |
summary | 1 factual sentence; 2 only if necessary |
tags | 2-5 keywords, such as names, places, topics, or goals |
importance | high, medium, or low; default is medium |
state/observation.txt and recent state files.done.state/memory.jsonl before writing.state/memory.jsonl.done.Run forgetting periodically through execute_skill with:
{"memory_file":"state/memory.jsonl","current_tick":100}
Maintenance may set _retention and _faded, and may prune weak lines.
Tune via environment variables described in references/forgetting.md.
importance: high only for facts that should persist across the simulation.Self-check (optional): python scripts/validate_memory.py state/memory.jsonl
For details, use:
references/memory_policy.mdreferences/forgetting.mdreferences/examples.md