agent-daily-paper
PassAudited by ClawScan on May 1, 2026.
Overview
This skill appears aligned with creating arXiv paper digests, but it uses local setup scripts, external package/model downloads, optional OpenAI credentials, and recurring scheduled pushes that users should review before enabling.
Before installing, verify that you trust the source, run the bootstrap in an isolated conda environment, decide whether to use local Argos or OpenAI translation, review any cron/automation command and delivery destination, and keep the config/output directories private if your research interests or delivery IDs are sensitive.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Running the bootstrap can download and install external Python packages and models into a local environment.
The bootstrap installs third-party packages without pinned versions and also preloads translation/embedding models. This is expected for the stated local paper-digest workflow, but users should trust the source and dependency chain before running it.
run([conda, "run", "-n", env_name, "python", "-m", "pip", "install", "argostranslate"], cwd=root) ... run([conda, "run", "-n", env_name, "python", "-m", "pip", "install", "sentence-transformers"], cwd=root)
Run setup in an isolated environment, review the source and dependency versions, and pin or verify packages/models if you need reproducible or higher-assurance installation.
If enabled, the digest can continue running on a schedule and posting or saving results until the scheduled task is removed.
The skill can create recurring automation for scheduled paper pushes. This is central to the stated daily-digest purpose and appears user-directed, but it is persistent background behavior.
如果用户要求定时推送，优先创建“精确到点”的 cron / 自动任务
Confirm the cron/automation time, timezone, command, and destination before enabling it, and keep a note of how to disable or remove the scheduled task.
If OpenAI translation is enabled, the configured API key may be used and paper text may be sent to the selected OpenAI model.
OpenAI credentials are optional and disclosed for translation. No artifact evidence shows hardcoded keys, logging, or unrelated credential use.
翻译提供方 `TRANSLATE_PROVIDER`：`argos`(默认) / `openai`(需 `OPENAI_API_KEY` + `OPENAI_TRANSLATE_MODEL`) / `auto` / `none`
Use the default local Argos mode if you do not want external API use; if using OpenAI, provide a limited key and verify the model and billing expectations.
Your research-interest configuration, local seed corpus, embeddings, and digest state may remain on disk and affect future recommendations.
The skill persists field-related seed documents, embeddings, and cached field fingerprints for reuse. This supports ranking and avoids repeated downloads, but persistent local context can influence later digests.
`prepare_fields.py` 会在每个领域落盘 Top-K 种子语料：`output/seed_corpus/docs/<canonical_en>.md` ... `output/seed_corpus/embeddings/<canonical_en>.json` ... 默认启用“领域指纹缓存”
Keep the config and output directories private, review field profiles before reuse, and delete or refresh cached seed/state files if they become stale or untrusted.