Skillv1.0.1

ClawScan security

agent-daily-paper · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 17, 2026, 1:00 PM

Verdict: Benign
Confidence: high
Model: gpt-5-mini
Summary: This skill is internally consistent with its stated purpose (arXiv-based daily paper digests) — it contains scripts to bootstrap an environment, fetch arXiv entries, run local ranking/translation, and emit Markdown; it does not request unrelated credentials or contain obvious exfiltration endpoints.
Guidance: This repo appears to implement exactly what it claims: a local arXiv digest pipeline that downloads models, creates a conda environment, builds seed corpora, and schedules runs. Before installing or enabling scheduled runs you should: 1) Review scripts (especially bootstrap_env.py, run_digest.py, and the cron/automation templates) to confirm paths and substitutions (USER_HOME, NODE_VERSION) are correct for your machine. 2) Run doctor.py first to validate config/subscriptions.json and dependencies. 3) Be prepared for large downloads (sentence-transformers/Hugging Face models, Argos packages) and run them in an isolated environment or VM if you have concerns. 4) Only provide OPENAI_API_KEY or any Feishu/third-party credentials if you intend to enable those integrations; the skill will function offline with Argos but OpenAI is optional. 5) The scripts fetch arXiv via the public API and write files under the repo (config/, data/, output/); back up or sandbox as needed. If you want extra assurance, run the bootstrap/doctor steps manually rather than granting full autonomous scheduling until you confirm behavior.

Review Dimensions

Purpose & Capability: okName/description (arXiv daily digest, bilingual cards, scheduling) match the repository contents and required actions: preparing field profiles, fetching from arXiv API, local ranking/embedding, optional translation, and producing Markdown output. Required files and scripts (prepare_fields.py, run_digest.py, bootstrap_env.py, etc.) are appropriate for the described functionality.
Instruction Scope: noteSKILL.md and scripts instruct the agent to run repository-local scripts, create/initialize config files, write outputs under output/daily, and create cron/autonomy templates in the user's home workspace. That is within the scope of a scheduling/digest skill, but users should note the templates require substitution of USER_HOME/NODE_VERSION and the scripts will create and overwrite config/subscriptions.json, state files, and seed corpora. SKILL.md and scripts also reference optional delivery templates (e.g., Feishu) and require a proper setup_required flow before automated pushes — these are expected but worth confirming before enabling automation.
Install Mechanism: noteThe skill is instruction-only (no platform install spec) but includes bootstrap scripts that run conda/pip and invoke model downloads (Argos packages, sentence-transformers / Hugging Face models, BAAI models). These operations will fetch third-party packages and model artifacts at runtime and write them to disk. This is expected for a local NLP pipeline, but users should be aware of large network downloads and that code/model artifacts come from package/model indexes rather than a vetted bundle.
Credentials: noteThe skill does not require any environment variables by default. It does, however, optionally read TRANSLATE_PROVIDER and OPENAI_API_KEY (for using OpenAI-based translation/reranking) and suggests delivery placeholders such as FEISHU_USER_ID in templates. These optional env vars are proportional to the optional features, but review any credentials you supply and only provide them when you intend to use OpenAI or external delivery integrations.
Persistence & Privilege: okalways is false and the skill does not request permanent platform-level privileges. It writes its own config/state under the repository (config/, data/, output/) and recommends creating user cron/automation jobs in the user's home workspace; this is normal for a scheduled digest tool and does not modify other skills or global agent settings.