Clawhub Memory Tiers Pro
v1.0.0Production-grade agent memory and quality system for multi-agent swarms. Extends agent-memory-tiers with an 8-point quality grading rubric, progressive discl...
⭐ 0· 102·0 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
Name and description promise a quality/rubric/coordination layer for multi-agent swarms; the SKILL.md contains an 8-point rubric, progressive-disclosure tiers, orchestrator patterns, and scoring templates — all coherent with that purpose. No unrelated binaries, env vars, or installs are requested.
Instruction Scope
The instructions explicitly direct auditors/orchestrators to read agents' SOUL.md and workspace content to score and build logs. That is expected for a swarm-quality skill, but it implies cross-agent data access and aggregation; users should be aware this will surface each agent's configuration/content (ensure no secrets are stored in SOUL.md or workspace files).
Install Mechanism
Instruction-only skill with no install spec and no code files. This minimizes filesystem and network risk; nothing will be downloaded or written by an installer.
Credentials
No environment variables, credentials, or config paths are requested. The skill's functionality (reading agent docs, scoring, templates) does not require external secrets, so requested access is proportionate.
Persistence & Privilege
The skill is not always-on and is user-invocable; model invocation is allowed (platform default). This is appropriate for the skill's purpose. Be mindful that autonomous agent invocation could cause the orchestrator to read or modify many agents if the orchestration layer grants those rights — manage orchestrator permissions accordingly.
Assessment
This skill appears to do what it claims and requests nothing unexpected, but before installing: 1) ensure the required dependency (agent-memory-tiers) is the trusted implementation you expect; 2) confirm agents' SOUL.md and workspace files do not contain secrets or credentials (the skill will read those docs to score agents); 3) run it first in a test swarm or sandbox to observe what it reads/writes and what outputs it generates; 4) restrict the orchestrator/agent permissions so it can only access the workspaces you intend; and 5) monitor outputs and logs and revoke the skill if it begins to access unrelated data. If you want more assurance, provide the full SKILL.md (untruncated) so I can re-check for any hidden instructions that reference external endpoints or env vars.Like a lobster shell, security has layers — review code before you run it.
latestvk97d5zd9amrevf5kdkwgepr5wd832wa4
License
MIT-0
Free to use, modify, and redistribute. No attribution required.