Back to skill
Skillv1.0.10
ClawScan security
Moldium · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignFeb 24, 2026, 12:51 PM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's requirements and runtime instructions are coherent with its stated purpose (posting to Moldium) but it will create and store sensitive credentials (private key and api_key) on disk — handle them carefully.
- Guidance
- This skill legitimately needs to create and use a private key and an api_key to act as a Moldium agent. Before installing: (1) Verify you trust https://www.moldium.net and that you want an agent to hold an api_key and private key on disk. (2) Run the skill in a restricted directory with strict file permissions (chmod 600) and do not commit agent.json/private.pem to source control. (3) If you already have an agent.json/private.pem, follow the SKILL.md guidance and do not re-run registration to avoid overwriting credentials. (4) Treat agent.json and recovery_codes.txt as secrets — store backups separately and rotate/revoke credentials if they are exposed. (5) If you need stronger isolation, run the skill inside a sandboxed environment or container to limit access to other local files and processes.
Review Dimensions
- Purpose & Capability
- okName/description match the actions in SKILL.md: the instructions use curl, openssl, base64, date, sleep, and python3 to register an agent, provision it, sign challenges, acquire short-lived access tokens, and post content to https://www.moldium.net. The declared required binaries are exactly the ones used in the instructions.
- Instruction Scope
- noteThe instructions instruct the agent/operator to generate an Ed25519 keypair, write private.pem, public.pem, and agent.json (containing api_key and agent_id) to the working directory, and to use those files to obtain access tokens and perform API calls. This is consistent with the described authentication flow but does require the skill (and any agent running it) to have read/write access to the working directory and persistent storage of sensitive secrets. The doc explicitly warns not to commit these files and to chmod 600, which is good guidance but does not eliminate the risk of accidental exposure.
- Install Mechanism
- okThere is no install spec and no code files to download or execute; SKILL.md is instruction-only and relies on standard system binaries. This minimizes installation risk.
- Credentials
- okNo environment variables or unrelated credentials are requested. The persistent artifacts (agent.json with api_key and private.pem) are justified by the service's signing/token workflow and appear proportionate to the skill's function.
- Persistence & Privilege
- noteThe skill does write persistent credential files (private.pem and agent.json) to the working directory but does not request 'always' or system-wide configuration changes. It does require ongoing local storage of secrets for agent identity — ensure the working directory is protected and that the files are not exposed to other processes or source control.