Skillv1.0.0

ClawScan security

油价智能提醒 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 27, 2026, 6:01 PM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's code, instructions, and requirements are consistent with an oil-price reminder: it fetches the official NDRC site, derives a trend, writes a local state file, and prints reminder text — there are no unexplained credentials, external endpoints, or install actions.
Guidance: This skill appears coherent and performs only the expected actions (fetch the NDRC site, compute a trend, write a local JSON, and output a reminder). Before installing: (1) confirm you are comfortable with the agent creating a cron job and writing a state file at /workspace/memory/oil_state.json; (2) verify how the platform's 'message tool' will deliver reminders (the script itself only prints the message — the platform must handle pushing it to you); (3) be aware the script uses a hard-coded base date to estimate the next adjustment window (a correctness/robustness issue you may want fixed); and (4) avoid overly frequent checks to respect the source site (SKILL.md already recommends ≤1/day). If you want stronger guarantees, ask the author to (a) derive the last adjustment date from the site or stored state instead of a hard-coded date, and (b) have the script call the messaging API explicitly (or document exactly how the platform will route printed messages to you).

Review Dimensions

Purpose & Capability: okName/description match the implementation: the script fetches ndrc.gov.cn, looks for '油价' announcements, derives 'up'/'down'/unknown and writes a local state file. Declared dependencies (Python, curl, cron, messaging) align with the described behavior and are proportionate to the stated purpose.
Instruction Scope: noteSKILL.md instructs the agent to create a cron job and a memory file and to push messages via a 'message tool'. The included script itself only fetches pages, computes a next-adjust date, writes /workspace/memory/oil_state.json and prints a message — it does not create cron jobs or call a messaging API. This is an implementation/coordination difference (orchestration vs script) rather than hidden behavior. Also note a functional inconsistency: the script uses a hard-coded base date (datetime(2026, 3, 23)) to compute the next adjustment window rather than deriving the last adjustment date from site content or stored state as the prose implies; this is a correctness/robustness issue but not an evidence of malicious intent.
Install Mechanism: okNo install spec (instruction-only with an included script) — low risk. The script calls the system curl via subprocess.run (no shell execution), does not download or execute arbitrary remote code, and writes only a local JSON state file. No remote install URLs or archive extraction are present.
Credentials: okThe skill requests no environment variables, no credentials, and no config paths beyond writing a state file under /workspace/memory — all proportional to a scheduling/monitoring reminder skill. External network access is limited to the official ndrc.gov.cn site (expected for this purpose).
Persistence & Privilege: noteThe skill writes a persistent local state file in /workspace/memory/oil_state.json and SKILL.md expects the agent to create a cron job to schedule runs. Those are reasonable for a periodic reminder skill, but users should confirm that the agent's cron creation will be scoped to the agent's workspace and not modify unrelated system cron entries. always:false and no elevated privileges are requested.