OpenCortex

v3.6.7

Self-improving memory architecture for OpenClaw agents. Structured memory files, nightly distillation, weekly synthesis, enforced principles (P0 for custom,...

2· 1.5k·7 current·8 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
Capability signals
Crypto
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The name/description (self-improving memory, distillation, synthesis, vault, git backup) matches the provided scripts and SKILL.md. Required binaries (grep, sed, find) and optional tools (git, gpg, openclaw) are appropriate for the declared tasks. No unrelated credentials or tools are requested.
Instruction Scope
The SKILL.md and scripts instruct the agent to read and write many workspace files, register cron jobs, and optionally build voice profiles or collect infra details. Those actions are consistent with the memory-management purpose, but they do grant the skill read/write access to essentially the entire workspace and autonomous cron-based activity — pay attention to opt-in toggles (VOICE_PROFILE, INFRA_COLLECT, SCRUB_ALL) before enabling them.
Install Mechanism
No external download/install spec is present; all provided code is plain bash in the skill bundle. The installer claims and appears to make no network calls during setup. Updates are performed by user-invoked 'clawhub install --force' (external download) which is a normal update flow and requires explicit action.
Credentials
No required environment credentials are declared. Optional env vars (workspace, timezone, vault passphrase, feature flags) are reasonable and documented; the only sensitive env is OPENCORTEX_VAULT_PASS which is justified for the encrypted vault. The skill advises preferring system keyring over file-based passphrases.
Persistence & Privilege
The skill registers two cron jobs that run isolated OpenClaw agent sessions scoped to the workspace. Autonomous cron execution is necessary for nightly distillation/weekly synthesis and is expected, but it does mean the skill has ongoing, automated execution capability; 'always:true' is NOT set and it does not modify other skills' configs.
Assessment
OpenCortex is internally consistent with its stated purpose, but review and decide on opt-in features before installing. Recommended steps: (1) run the installer in dry-run mode first (bash skills/opencortex/scripts/install.sh --dry-run); (2) inspect scripts/vault.sh and scripts/git-backup.sh to ensure their behaviors align with your policies; (3) keep .vault/ and .secrets-map in .gitignore as the installer suggests; (4) avoid enabling voice profiling or infrastructure collection unless you are comfortable with the agent reading conversation logs and infra mentions (these are off by default and require explicit flags); (5) prefer system keyring to file-based passphrases (do not set OPENCORTEX_ALLOW_FILE_PASSPHRASE unless necessary); (6) if you use git backup, avoid passing --push unless you intend to push scrubbed commits to a remote; (7) after install, run bash skills/opencortex/scripts/verify.sh and review the cron entries (openclaw cron list / crontab -l) to confirm scheduled jobs; (8) updates require an explicit clawhub install --force (no silent network pull during initial install).

Like a lobster shell, security has layers — review code before you run it.

architecturevk97c8s8erk25ak9j6carmhsztx81p5c6latestvk976kbt1t5bw37jcm1a4099g9x84q1tjmemoryvk97c8s8erk25ak9j6carmhsztx81p5c6self-improvingvk97c8s8erk25ak9j6carmhsztx81p5c6voicevk9734b9c7a1n8b1dvbndafmgkd81mjb1

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Runtime requirements

Binsgrep, sed, find

Comments