skills.rest
awebai
awebaiOFFICIAL
Explore all skills by this author →
GitHub

self-maintenance

Records durable knowledge into agent docs, decisions, memory, and skills directories.

82|9|Updated Feb 2, 2026
Search Tags:#documentation#markdown#decisions#session continuity#agent memory#self-maintenance#knowledge persistence
One-click install
npx skills add https://github.com/awebai/aweb --skill self-maintenance-awebai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: self-maintenance
Source: https://github.com/awebai/aweb/tree/main/.agents/skills/self-maintenance
Command: npx skills add https://github.com/awebai/aweb --skill self-maintenance-awebai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents lose everything they learn when a session ends, forcing each new session to rediscover the same facts. This Skill defines how an agent persists durable knowledge into its soul directory so future sessions of the same role keep what earlier sessions learned. ## Core Features & Use Cases - Structured Knowledge Growth: Writes durable findings into docs, decisions, memory (indexed in MEMORY.md), and reusable skills under the agent's role directory. - Identity Protection: Enforces a hard rule that AGENTS.md and role definitions are never edited mid-work, keeping identity human-owned while knowledge grows. - Quality Bar: Applies a two-part test (durable and behavior-changing) before recording anything, preventing noisy, rotting memory files. - Use Case: While debugging, an agent discovers a non-obvious deployment constraint. Instead of leaving it in chat, it writes a memory entry and updates MEMORY.md so the next session starts with that fact. ## Quick Start Use the self-maintenance skill to record this durable finding into my agent memory and update the index.

Frequently Asked Questions about self-maintenance

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make an AI agent remember things between sessions?▼

Store durable facts as markdown files in a persistent directory such as agents/<role>/memory/, indexed in a MEMORY.md file. Each new session reads these files at startup, so knowledge survives session restarts without relying on chat history.

What should an agent write to its long-term memory?▼

Record something only when it is durable and would change what a future session does, such as non-obvious constraints, committed decisions, or recurring procedures. Skip routine work logs, anything already in code or git history, and prune entries that go stale.

Can an agent modify its own AGENTS.md or role definition?▼

No. This Skill explicitly forbids editing AGENTS.md or role definitions because identity is human- and review-owned. Agents grow their knowledge files only; identity changes go through deliberate human review.

What should never be stored in agent memory files?▼

Never store secrets, invite tokens, private keys, .aw state, DIDs, addresses, or certificates in soul or memory files. These directories are committed to the repo through the normal review path, so sensitive material would be exposed.

When should an agent create a new skill instead of a memory entry?▼

Create a skill under .agents/skills/<name>/SKILL.md when a real, recurring procedure appears that you will need again. Single durable facts belong in memory files; reusable multi-step procedures belong in skills.

Related Skills

Explore More Software Engineering Skills→
ignidusignidus

docs-generation

Convert a knowledge graph into a structured docs/ folder with TOC and AGENTS.md.

Official
Advanced
TyrizxTyrizx

merge

Merge a topic branch into the base branch and resolve conflicts in the main worktree.

Official
Intermediate
Shockwave2kShockwave2k

consolidate-memory

Merge duplicate memory topics and prune the MEMORY.md index.

Community
Intermediate
Cloud-Byte-ConsultingCloud-Byte-Consulting

wiki-synthesis-agent

Synthesize source material into maintained wiki pages with citations and contradiction flags.

Official
Intermediate