Skillv1.0.0

ClawScan security

mar-dreaming · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

BenignApr 24, 2026, 3:40 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's files, runtime instructions, and required tools match its stated purpose (periodic, local creative output); nothing requests unrelated credentials or network access.
Guidance: This skill appears to do exactly what it says: locally decide when to produce short creative journal entries and store them under data/ and memory/dreams/. Before installing: ensure jq and (optionally) python3 are installed; confirm WORKSPACE is unset or points to the intended project directory so files aren't written elsewhere; inspect or back up any existing data/dream-state.json or memory/dreams to avoid accidental overwrite; be aware the script updates data/dream-state.json each run (concurrent heartbeats could race). If you want stricter safety, run the script in a container or from a workspace copy first to validate behavior.

Purpose & Capability: okName/description (creative background dreaming) align with what is included: a small gate script that decides whether to 'dream' and JSON/files for state and outputs. Required binaries (jq, python3) are appropriate for JSON parsing and randomness used by the script.
Instruction Scope: okSKILL.md instructs the agent only to run the included script during quiet-heartbeat checks and to write output to local memory/dreams files. The instructions do not ask the agent to read unrelated system files, network endpoints, or secrets.
Install Mechanism: okNo install spec; this is instruction-only with a small included shell script and state/config JSON. Nothing is downloaded or executed from remote URLs and no archives are extracted.
Credentials: noteNo credentials are requested. The only env var mentioned is an optional WORKSPACE that changes where data and memory are written — useful and expected, but if set to an unexpected path it could cause the script to write outside the intended workspace. Recommend verifying WORKSPACE before use.
Persistence & Privilege: okThe skill is not always-on and is user-invocable; it does not request broad platform privileges, nor does it alter other skills' configs. Autonomous invocation remains possible (platform default) but is not combined with other red flags.