Menú
Defines how to save and load agent state as checkpoint files.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
| name | memory-checkpoint |
| type | instruction |
| description | Defines how to save and load agent state as checkpoint files. |
Defines how to save a checkpoint and how to load one into SI. When to checkpoint is determined by the calling context — the project, the war room, or the agent's own judgment.
Create a timestamped folder under memory/checkpoints/. Inside it, write one file per agent indexed by agent number.
memory/checkpoints/YYYYMMDD_HHMMSS/
├── agent-0.md
├── agent-1.md
└── ...
Each agent-N.md captures that agent's state. No length constraints — summarise what is needed to resume. A good checkpoint captures what would be hard to reconstruct, not a verbatim log.
# Agent <N> — <timestamp>
**Role**: <role>
**Trigger**: <what caused this checkpoint>
**Completed**: <what was just finished>
**Next**: <exact next action>
## State
<key facts, file paths, env state, decisions made, blockers — whatever is needed to resume>
After writing all agent files, update memory/SUMMARY.md — rewrite it to point to the latest checkpoint.
# Summary
- **Last checkpoint**: `memory/checkpoints/<timestamp>/`
- **Status**: <not_started | in_progress | blocked | complete>
## State
<key facts — file paths, important values, decisions that affect what comes next>
## Resume From
<one clear instruction: what to do next, any preconditions>
On resume, inject into SI in this order:
memory/SUMMARY.md — always first. Get current status and the latest checkpoint path.agent-N.md from the latest checkpoint folder — inject the file for your agent index.memory/
├── SUMMARY.md
└── checkpoints/
├── 20260328_090000/
│ └── agent-0.md
├── 20260328_142301/
│ ├── agent-0.md
│ └── agent-1.md
└── 20260328_154732/
├── agent-0.md
└── agent-1.md
Mechanism for keeping an agent anchored to a caller-defined persona across sessions. Covers the biplate private memory system for drift detection and self-correction.
Defines the structure and requirements for a research proposal.
Multi-agent research war room. Personas debate in sequential turns through two phases — ideation and proposal writing. Persona persistence and drift detection are enforced every turn via the persistent-persona skill.
Search and summarize papers from ArXiv. Use when the user asks for the latest research, specific topics on ArXiv, or a daily summary of AI papers.
Core research principles that govern all project decisions. Load this skill for every research project. Every idea, method, and artifact must be evaluated against these principles before proceeding.
Probe system resources — CPU, RAM, disk, GPU/CUDA/MPS — and translate findings into experiment design recommendations. Run this at the start of any compute-intensive project.