Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add jiankang1991/nsfc-benzi-audit --skill nsfc-benzi-auditgit clone --depth 1 https://github.com/jiankang1991/nsfc-benzi-auditWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/jiankang1991/nsfc-benzi-audit/nsfc-benzi-audit)<a href="https://agentmods.dev/skills/jiankang1991/nsfc-benzi-audit/nsfc-benzi-audit"><img src="https://agentmods.dev/badge/skills/jiankang1991/nsfc-benzi-audit/nsfc-benzi-audit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/jiankang1991/nsfc-benzi-audit/nsfc-benzi-audit"><img src="https://agentmods.dev/badge/skills/jiankang1991/nsfc-benzi-audit/nsfc-benzi-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Anti-Refusal · line 30 Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.Fix: Remove instructions that suppress warnings, disclaimers, or ethical commentary. Let the agent surface safety-relevant caveats to the user.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00154 | $0.02063 |
| Opus 5.5 | $0.00062 | $0.00825 |
| Sonnet 5.5 | $0.00031 | $0.00413 |
| Haiku 4.5 | $0.00015 | $0.00206 |
Grade A, and why
nsfc-benzi-audit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 19d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
NSFC Benzi Audit
Produce applicant-facing diagnosis and revision advice grounded in the supplied draft. Expose substantive logic breaks, weak evidence and high-impact fixes. Do not fabricate literature, results, project histories, budgets or official requirements.
This workflow gives applicant revision advice. If the user requests formal communication review, identify that separate scope and use a suitable review skill only if available. No companion skill is required for draft-internal diagnosis.
Workflow
- Select the source and establish the review scope.
- Prioritize the source and revision explicitly identified by the user. Separate target drafts, funded examples, old reports, templates and notes. Never select a draft merely because it is the largest Markdown file or named
full.md/output.md. - Reuse extracted text only after matching its title, revision and contents to the selected original. Record source path/version and extraction provenance. If multiple plausible drafts cannot be distinguished, ask which one to use.
- For PDF/DOCX, use available extraction/OCR tools and preserve headings and tables. If extraction is unavailable, state the limitation and work on readable supplied material; do not invent a tool or require a named companion skill.
- Inspect original figures when judging visual logic or evidence, including preliminary results. Converted mermaid/OCR is not authoritative. Missing extraction is a material gap, not proof that the applicant omitted content.
- Record application year, exact project category (including 青年 A/B/C where known), form version/source, funding mode (包干制/预算制/待确认), research attribute, application code and source completeness. Keep historical-example years separate from the target year.
- Use questions, goals and methods as analytical dimensions wherever they occur. Require separate headings only when the applicable form requires them. If year or form is unknown, continue content diagnosis and mark rule applicability unresolved.
- For a quick pass, focus on title, abstract, scientific questions, contents, innovation and basis. For full diagnosis, cover all supplied sections. A 专项 request (one chapter, one question) stays within that scope. Do not report absent attachments or omitted excerpts as confirmed defects in the original application.
- For long or multi-file drafts, keep section/page locators and a running list of candidate findings with their locations while reading; before closing, revisit only unresolved candidates and report any source ranges you could not read.
- Prioritize the source and revision explicitly identified by the user. Separate target drafts, funded examples, old reports, templates and notes. Never select a draft merely because it is the largest Markdown file or named
What ships with it
14 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 273 B
- assets/report-template.md 5.2 KB
- LICENSE 1.1 KB
- references/audit-surfaces.md 20 KB
- references/benzi-logic.md 43 KB
- references/current-rules.md 8.3 KB
- references/exemplar-learning.md 8.2 KB
- references/geospatial-remote-sensing.md 13 KB
- references/information-communication.md 8.9 KB
- references/kd-lookup.md 16 KB
- references/medical-biomedical.md 14 KB
- references/question-distillation.md 16 KB
- references/representative-works.md 6.6 KB
- scripts/nsfc_ack_grants.py 4.5 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 19d ago Changed ebcc6bf1278c
- 1mo ago Changed 078eb37a84cd
- 1mo ago Changed · +1 lines · +58 tokens per session d046081fbcb3
- 1mo ago Changed · -18 lines · -101 tokens per session a36cad1326bf
- 2mo ago Changed 60fdca811c36
- 2mo ago First seen · 77 lines · 197 tokens per session scan A 7ac56e9662b8
nsfc-benzi-audit is a skill published in the GitHub repository jiankang1991/nsfc-benzi-audit (131 stars, last pushed 8d ago), licensed MIT. It adds 154 tokens to every session and 2,063 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
nsfc-benzi-audit compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| nsfc-benzi-auditjiankang1991/nsfc-benzi-audit | Skill | 131 | 8d ago | MIT | A | 154 | Codex |
| patent-maphandsomestWei/patent-disclosure-skill | Skill | 13k | 2d ago | MIT | A | 48 | Claude Code |
| autoreviewopenclaw/clawhub | Skill | 9.5k | yesterday | MIT | A | 17 | Claude Code, Codex |
| sandbaseiflytek/skillhub | Skill | 5.4k | yesterday | Apache-2.0 | A | 67 | Claude Code |
| convex-insightsopenclaw/clawhub | Skill | 9.5k | yesterday | MIT | A | 45 | Claude Code, Codex |
| openclaw-design-systemopenclaw/clawhub | Skill | 9.5k | yesterday | MIT | A | 37 | Claude Code, Codex |
| convex-addopenclaw/clawhub | Skill | 9.5k | yesterday | MIT | A | 81 | Claude Code, Codex |