HumanLike Memory / Human-Like Memory is a persistent AI agent memory system for long-term memory search, recall, and save. It helps agents store facts, prefe...
Security Analysis
high confidenceThe skill's code, instructions, and requested credentials are consistent with a human-like long-term memory service and do not ask for unrelated secrets or system access.
Name/description, declared primary credential (HUMAN_LIKE_MEM_API_KEY), required OpenClaw config path, and the node runtime requirement all align with a memory service client. The CLI commands and documented behavior match the stated purpose.
SKILL.md and scripts instruct only explicit recall/search/save operations and only send data when invoked. The documentation and code explicitly state they do not read ~/.openclaw/secrets.json, arbitrary files, or other environment variables; the code enforces an allowlist of HUMAN_LIKE_MEM_* env vars.
No install script or remote download is present; the skill is distributed with its source files and is instruction-only for installation. The only runtime dependency is node (>=18), which is reasonable for the shipped .mjs scripts.
The skill uses a single primary credential (HUMAN_LIKE_MEM_API_KEY) and a small set of HUMAN_LIKE_MEM_* env vars — appropriate for this purpose. Minor metadata inconsistency: registry metadata lists 'Required env vars: none' while primaryEnv is set to HUMAN_LIKE_MEM_API_KEY and config path skills.entries.human-like-memory.apiKey is required; this is a documentation/metadata mismatch rather than a functional problem.
always is false and the skill does not request forced/global presence. It does not modify other skills or system-wide config beyond recommending OpenClaw config set commands (which are standard and executed only if the user/agent performs them).
Guidance
This skill appears to implement a straightforward memory client. Before installing, verify you trust the remote endpoint (https://plugin.human-like.me) or configure a self-hosted base URL if you prefer. Only explicit save/retrieval operations are sent to the service, but avoid sending passwords or other secrets to the memory server. Note the small metadata mismatch: the registry lists no required env vars while the skill expects HUMAN_LIKE_MEM_API_KEY (or the equivalent OpenClaw config entry); ensure you set the API key via openclaw config yourself rather than having an agent invent or fetch secrets. If you need stricter privacy, self-host the backend and set HUMAN_LIKE_MEM_BASE_URL to a URL under your control.
Latest Release
v1.0.12
Backport procedural memory v2 support to openclaw-skill, add explicit scenario to writes, and keep v1 fallback for compatibility.
More by @humanlike2026
Published by @humanlike2026 on ClawHub
