agentleFS
longkou1988longkou1988/cnki-skills/skills/cnki-jev/SKILL.md

cnki-jev

Skill28 starsChanged 2 months agoMITView on GitHub

Optional paid Jev-assisted CNKI title/abstract screening with per-condition proposals, budget accounting, and escalation to evidence-based LLM or human review. Use when the user chooses Jev screening or a shadow comparison.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read configuration and workflow.

    line 14
  2. Very specific:

    Start with the supplied disabled config.

    line 22
  3. Very specific:

    Keep the task and its adjacent .jev.sqlite together across retries/resumes.

    line 25
  4. Distinctive:

    Run the helper for each eligible paper, reusing the same task and configuration.

    line 32
  5. Distinctive:

    Only title and abstract are sent.

    line 33
  6. Very specific:

    Read the actual source and save the final include/exclude/uncertain decision, criterion-linked reason and exact evidence using the existing …

    line 40

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CNKI Jev 辅助筛选(可选,调用付费)

Requires cnki-screening and cnki-resume. The base bundle works without this skill. Installation never enables calls. The helper uses Python 3.9+ standard library and TypeSafe's official hosted API; it does not browse CNKI or read/upload full text.

Before enabling

Read configuration and workflow. Preserve existing task criteria and authorized scope. User authorization must cover paid calls, a budget, and transmission of the task's titles/abstracts to TypeSafe. If these were already specified, use them without asking again. Otherwise obtain only the missing choices. The user configures TYPESAFE_API_KEY in their environment; never ask them to paste it into a conversation, config file, task ledger or repository. Do not purchase credits or create an account as an implicit setup step.

Start with the supplied disabled config. shadow and assist both incur fees; off does not. State that the local USD budget is an estimate based on a configured rate, not a provider-enforced billing cap; the call-count limit is enforced locally. Keep the task and its adjacent .jev.sqlite together across retries/resumes.

Screening

  1. Freeze/import the actual result set with cnki-resume. Match criteria exactly and decompose all necessary inclusion conditions into atomic questions. Check coverage against the research question; do not silently drop/change conditions.
  2. Run the helper for each eligible paper, reusing the same task and configuration. Only title and abstract are sent. Missing abstracts return to baseline review.
  3. shadow: let the baseline LLM/human make an independent decision before comparing the saved Jev proposal. Do not use Jev's suggestion as the evaluation reference. assist: an inclusion proposal proceeds to source-evidence verification; exclusions and uncertain/complex/conflicting cases go to baseline LLM/human review. A high probability is not demonstrated accuracy. The supplied threshold is an uncalibrated starting value, not permission for automatic exclusion.
  4. Read the actual source and save the final include/exclude/uncertain decision, criterion-linked reason and exact evidence using the existing screen command. Include the proposal provenance in that record (see reference). Never submit a proposal as a final screening record, fabricate a quote, or treat missing text as a failed criterion. Full-text reading, when needed, stays in the base workflow.
  5. Report candidates, final decisions, unresolved reviews, actual route, request count and estimated usage/reserved budget separately. Export the usual workbook plus local Jev proposals if requested. Never upload private ledgers, abstracts, proposal files or credentials to GitHub as part of publishing the skill.

On API/configuration/budget failure, honor on_failure: baseline means the agent continues ordinary screening; pause means preserve progress and stop that portion. The helper routes work but does not itself invoke a second model. It never retries an uncertain paid request automatically. See the reference for manual failed-call retries and crash recovery. Do not reset/delete the sidecar to evade a budget.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.