Agent Memory Tiers
v1.0.0Tiered memory system for OpenClaw agents. Gives agents instant context on startup with a 4-line state snapshot (L0) and 7-day rolling context (L1). Eliminate...
⭐ 0· 106·0 current·0 all-time
bySwarm Forge@dirtyrootsstudio
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (tiered agent memory) matches the SKILL.md instructions: create L0.md and L1.md, add headers/footers to SOUL.md, read L0/L1 at startup and update them at end-of-run. It does not request unrelated credentials, binaries, or installs.
Instruction Scope
Instructions are focused on reading/writing files in the agent workspace (L0.md, L1.md, SOUL.md) which is appropriate. They require the agent to always read L0/L1 before other context and to update them at end-of-run. The SKILL.md assumes the agent will implement these updates but provides no code or enforcement — this is an operational/design gap rather than a security mismatch. The instructions also permit reading memory/ only when needed; that is reasonable but grants the agent discretion which could broaden file reads if misused.
Install Mechanism
No install spec or code files are provided; this is instruction-only so nothing is written to disk by the skill itself. That is the lowest-risk install posture and consistent with the declared metadata.
Credentials
The skill requests no environment variables, credentials, or config paths. The only resource it uses is the agent workspace files it instructs to create and update, which is proportional to its purpose.
Persistence & Privilege
always:false and no install means it does not force permanent presence. The skill allows autonomous invocation (the platform default) but does not request elevated agent-wide privileges or modify other skills.
Assessment
This skill is a set of instructions, not runnable code — it tells agents to keep two small files (L0.md and L1.md) and to update them each run. Before using it: (1) confirm you’re comfortable with agents writing readable files in their workspace and ensure file permissions/backups as needed, (2) avoid placing secrets, credentials, or PII into L0/L1 (they are designed to be short, human-readable summaries), (3) test on a non-production agent to ensure the agent actually updates the files (the SKILL.md assumes the agent implements the update step), and (4) if you have strict data-handling requirements, add rules or automation to sanitize or encrypt any sensitive fields. If you want automatic updating, implement and review the agent-side code that performs the writes; the skill itself provides only the process, not the implementation.Like a lobster shell, security has layers — review code before you run it.
latestvk971pcd5bk3dkkn7z2jn5qgnt9832zq9
License
MIT-0
Free to use, modify, and redistribute. No attribution required.