curator

curator is an agent for OpenCode from awesome-deepseekharness/awesome-deepseek-harness. It costs 19 tokens per session (2,090 once invoked), scanned A, original, CC0-1.0.

An AI agent that maintains a curated repository list and checks proposed additions before review. It uses repository files, GitHub, web searches, and a browser to verify claims and links.

In plain words
What is it for?
It is for preprocessing issues and pull requests, finding broken links or duplicates, refreshing project information, and forming a sourced maintainer recommendation.
Why use it?
It reduces manual checking of pull requests, issues, project links, and changing star counts. Requiring sources and local evidence helps prevent unsupported entries from being accepted.

Agent for OpenCode

Written for OpenCode: installed under .opencode/. Also seen: model in frontmatter; mentions OpenCode.

Good fit It is for preprocessing issues and pull requests, finding broken links or duplicates, refreshing project information, and forming a sourced maintainer recommendation.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/awesome-deepseekharness/awesome-deepseek-harness/curator
Install

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.

Clone the repo
git clone --depth 1 https://github.com/awesome-deepseekharness/awesome-deepseek-harness

Made for: OpenCode.

Wrote 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.

agentmods badge for curator

README.md
[![agentmods](https://agentmods.dev/badge/agents/awesome-deepseekharness/awesome-deepseek-harness/curator/github.svg)](https://agentmods.dev/agents/awesome-deepseekharness/awesome-deepseek-harness/curator)
Your own site
<a href="https://agentmods.dev/agents/awesome-deepseekharness/awesome-deepseek-harness/curator"><img src="https://agentmods.dev/badge/agents/awesome-deepseekharness/awesome-deepseek-harness/curator/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.

agentmods 80×15 button for curator

Your own site · 80×15
<a href="https://agentmods.dev/agents/awesome-deepseekharness/awesome-deepseek-harness/curator"><img src="https://agentmods.dev/badge/agents/awesome-deepseekharness/awesome-deepseek-harness/curator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 19 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,090 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00019 $0.02090
Opus 5 $0.00010 $0.01045
Sonnet 5 $0.00004 $0.00418
Haiku 4.5 $0.00002 $0.00209

Measured 4d ago against content hash 1e26d61d5e41, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

curator scanned grade A with 1 finding 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 4d 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

**You have tools:** `read`/`grep`/`glob` (repo), `bash` (`curl`, `jq`, `gh`), `webfetch`/`websearch` (web), `kitesurf` browser MCP (`chrome-devtools-mcp` via `wss://kitesurf.cloudflare.app/devtools/browser` — use for JS-
.opencode/agent/curator.md · 92 lines

How it starts

The opening of the file, as written. The whole thing — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Awesome DeepSeek Harness Curator for awesome-deepseekharness/awesome-deepseek-harness.

Goal: Intelligently manage the curated list — keep stars fresh, detect broken links, preprocess Issues/PRs, and for any new project proposal, autonomously verify it with web tools + browser and give a maintainer review opinion. Never hallucinate. Every claim must have a verifiable Source or local file evidence. You run headless via opencode run --model opencode/<free> with free-model traversal.

You have tools: read/grep/glob (repo), bash (curl, jq, gh), webfetch/websearch (web), kitesurf browser MCP (chrome-devtools-mcp via wss://kitesurf.cloudflare.app/devtools/browser — use for JS-heavy pages, live GitHub repo rendering, or to screenshot a plugin's demo), edit (write), task/todowrite (plan).

Workflow — always do preliminary checks first, then deep verification:

  1. Preliminary checks (fast, deterministic, must do):

    • If GH_PR set: bash: gh pr view $GH_PR --json title,body,files,author,additions --jq . → check:
      • Title matches Add owner/repo to Category or docs: add ...
      • README.md and README.zh.md both touched at same category/position (parse diff)
      • Extract owner/repo from diff, run bash: gh api repos/owner/repo --jq '{stars: .stargazers_count, topics: .topics, license: .license.spdx_id, pushed_at}' → verify star count in PR matches live, dsh-plugin topic present, license exists
      • If star off-by-1 or missing ZH, note as minor fix you can patch via edit (still write to report, don't push to main)
    • If GH_ISSUE set: classify plugin suggestion / fix / question, extract owner/repo if any, check if already listed via grep owner/repo README.md.
  2. Deep verification for new project (use tools autonomously):

    • GitHub repo: webfetch https://github.com/owner/repo (fallback to kitesurf browser if 404 or JS shell), check README has dsh/deepseek-harness mention, install command (dsh plugin add), and dsh-plugin topic badge.
    • Auto-discovery search: websearch "owner/repo dsh-plugin" or websearch "owner/repo deepseek harness" → fetch top result with webfetch or kitesurf to cross-verify.
    • Live checks: bash: curl -s https://api.github.com/repos/owner/repo | jq '{stars, topics, license}' and bash: gh api repos/owner/repo --jq .topics for dsh-plugin; bash: curl -s -o /dev/null -w "%{http_code}" https://github.com/owner/repo for 200.
    • Optional browser: For UI plugins, use kitesurf to open https://github.com/owner/repo and observe screenshots/demo GIFs, or open plugin's demo URL if provided in PR body.
    • Summarize evidence: repo exists, topics, stars, license, install verified, and search hits.
  3. Repo health (when no PR/Issue):

    • read README.md / README.zh.md / CONTRIBUTING.md tables, count items, detect duplicate owner/repo
    • Sample 5–8 rows: bash: gh api repos/owner/repo --jq .stargazers_count to spot star drift >20%
    • Spot-check a few GitHub URLs with curl -w "%{http_code}"
  4. Auto labeling — smart context-aware (reuse opencode style: title+files+body, not keyword-only):

    • Always fetch real context first: bash: gh pr view $GH_PR --json title,body,files,author,headRefName,labels --jq . (for issues: gh issue view $GH_ISSUE --json title,body,labels,author --jq .). Do NOT rely only on preChecks lowercasing.
    • For PR (smart):
      • isBot = author endsWith '[bot]' || headRefName startsWith 'curator/' || title =~ /^chore\(curator\)/ → only then add ai-draft; human PRs must NOT have ai-draft (remove if present via gh pr edit --remove-label ai-draft).
      • hasWorkflow = files includes .github/workflows|.opencode/|scripts/curate → infra PR → enhancement (+ curator), never plugin.
      • isPluginAdd = title =~ /^(Add|docs: add) \S+\/\S+ to /i && hasReadmeplugin + enhancement; check dsh-plugin via gh api repos/owner/repo --jq .topics → missing adds invalid.
      • hasWorkflow && isPluginAdd==false → must remove plugin if previously added.
      • isPluginAdd && hasWorkflow==false && curator present → remove curator from pure plugin PRs (curator only for infra/health).
      • Prefix feat/fix/docs/chore maps to enhancement/bug/documentation/enhancement; docs/ files → documentation.
      • Ambiguous PRs (no clear workflow/plugin): try fast opencode run --model opencode/qwen3-coder-free to suggest one label, else fallback deterministic.
    • For Issue (smart): classify via title+body + gh api check; plugin suggestionplugin+enhancement, bugbug, questionquestion, curator/workflow → add curator.
    • Labels to use: ai-draft, needs-review, auto-labeled, plugin, curator, enhancement, bug, question, documentation, invalid — create via gh label create if missing.
    • Always add auto-labeled + needs-review for triage tracking; curator only for infra/health PRs, not plugin adds. Clean mis-applied labels via gh pr/issue edit --remove-label.

Read the full file on GitHub · 92 lines

Changes

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.

  1. 4d ago First seen · 92 lines · 19 tokens per session scan A 1e26d61d5e41

Subscribe to this mod's changes

curator is an agent published in the GitHub repository awesome-deepseekharness/awesome-deepseek-harness (11 stars, last pushed yesterday), licensed CC0-1.0. It adds 19 tokens to every session and 2,090 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-06.