← Back to Skills Marketplace
tiggy-chan

Anime Finder

by TiggyChan · GitHub ↗ · v1.4.0 · MIT-0
cross-platform ✓ Security Clean
100
Downloads
0
Stars
1
Active Installs
2
Versions
Install in OpenClaw
/install anime-finder
Description
Find anime releases and handle the whole request end to end. Use when the user asks to 找番、 追更、看最新一集、拿最新一季第一集、帮我下载、只要磁力、先找资源别下载、或者查看刚才下载怎么样; also use it for m...
README (SKILL.md)

Anime Finder

Prefer the single workflow entrypoint and pass the raw user request when possible:

python3 "$SKILL_DIR/scripts/workflow.py" "用户原话或番名" [--episode N] [--latest-season] [--download] [--prefer-4k] [--json]

find.sh still exists, but it is compatibility glue only.

Agent Workflow

  1. Pass the user utterance into workflow.py --json.
  2. Read top-level status, intent, and decision before digging into debug details.
  3. If decision.confirmation_required is true, ask at most one question.
  4. If intent.action is status, trust the workflow to read the last known download state; do not reconstruct it yourself.
  5. If the user explicitly wants only a magnet link, pass --downloader cli-only or let the raw utterance express it.

Stable Fields Worth Reading

  • status
  • intent
  • decision
  • best_result
  • verification.match / verification.alternatives
  • search.search_queries
  • search.available_seasons
  • search.resolved_latest_season
  • download.status
  • summary

Status Rules

  • need_disambiguation: stop and ask one focused question
  • no_results: search finished, but nothing downloadable remained
  • ready: best_result is available and no download was executed
  • queued: Transmission accepted the torrent or the last tracked task is still queryable
  • fallback_to_magnet: the workflow intentionally returned magnet/torrent output instead of queueing
  • blocked: local or network blockers prevented progress

Intent And Preference Rules

  • scripts/intent.py is the normalization layer. It extracts title, episode, season, latest_episode, latest_season, action, quality/subtitle/downloader preferences, confidence, and missing fields.
  • scripts/data/series_overrides.json remains the source of truth for franchises whose Bangumi names, user wording, and Nyaa season numbering drift apart.
  • Runtime user preferences live in state/user_profile.json. Only explicit user wording should write to it. The tracked template is state/user_profile.template.json.
  • Follow-up download queries read state/last_download.json; treat that file as runtime state, not repo content.

Example Requests

  • 帮我找一下 JOJO 最新一季第一集然后下载下来
  • jojo latest season ep1
  • 先找孤独摇滚第三集,别下载
  • 攻壳机动队只要磁力
  • 刚才那个下载怎么样

More trigger examples live in references/intent_examples.md.

Usage Guidance
This skill appears to do what it claims: it queries Bangumi and Nyaa, ranks releases, and can either return magnet links or push torrents to Transmission. Before installing/using: 1) Be aware it will perform network requests (api.bgm.tv, nyaa.si) and will write state files under the skill and may create ~/.transmission and files under your Downloads directory when downloading. 2) If you don't want automatic downloads, prefer using search-only or pass --downloader cli-only; review or edit state/user_profile.json to disable auto-download (auto_download_high_confidence). 3) Review the scripts (workflow.py, search_nyaa.py, verify.py) if you want to confirm behavior locally. 4) Consider legal/privacy implications of torrenting in your jurisdiction and run on a machine you control. The skill does not request secrets or external install scripts, which reduces supply-chain risk.
Capability Analysis
Type: OpenClaw Skill Name: anime-finder Version: 1.4.0 The anime-finder skill bundle is a well-structured tool for searching and downloading anime via Nyaa.si and the Bangumi API. It features a robust intent normalization layer (scripts/intent.py) and a central workflow (scripts/workflow.py) that manages identity verification and Transmission BitTorrent client integration. All high-risk operations, such as network requests to nyaa.si/bgm.tv and local execution of transmission-daemon, are strictly aligned with the stated purpose and include safeguards like non-anime intent filtering. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found.
Capability Assessment
Purpose & Capability
Name/description match implementation: scripts verify identity via Bangumi, search Nyaa (RSS), rank results, and optionally push torrents to a downloader (Transmission) or return magnet links. Required artifacts (state files, series overrides) are consistent with the described features.
Instruction Scope
SKILL.md instructs agents to run scripts/workflow.py with raw utterances and inspect the returned JSON. The workflow performs network calls (Bangumi API, Nyaa.si) and reads/writes local state (state/user_profile.json and state/last_download.json). It also can download .torrent files and invoke local commands to start/communicate with transmission-daemon. This behavior is coherent with the stated purpose but means the skill will perform network IO and modify local files and processes when used.
Install Mechanism
No install spec; this is instruction+source code shipped in the skill. There are no downloads from untrusted third-party URLs during install. Runtime network requests (Bangumi, Nyaa) are expected for search/verification.
Credentials
The skill requests no environment variables or external credentials. It does access the network and the filesystem (writes state files under the skill, default download dir $HOME/Downloads, and may create ~/.transmission when auto-starting Transmission) which is proportional to a downloader-type tool.
Persistence & Privilege
always:false (normal). The skill persists local state in state/*.json inside the skill and may create ~/.transmission and write downloaded .torrent files or initiate downloads into the user's Downloads directory. This is expected for a download-capable skill but is a change to local system state that users should be aware of.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install anime-finder
  3. After installation, invoke the skill by name or use /anime-finder
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.4.0
v1.4.0 - 意图识别层、决策合约、用户偏好持久化
v1.2.0
- 增加番剧身份确认流程,先用 Bangumi 精准识别番名后再搜索资源,避免误搜。 - 优化复杂季数映射逻辑,提升对季/部分/集数的识别准确性。 - 支持用户明确选择匹配番剧,模糊或多义名时自动列出备选供确认。 - 增强错误处理与用户友好提示,未找到资源或集数可智能说明原因和建议。 - 新增进阶使用指南与命令行用法速查,体验更清晰易用。
Metadata
Slug anime-finder
Version 1.4.0
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 2
Frequently Asked Questions

What is Anime Finder?

Find anime releases and handle the whole request end to end. Use when the user asks to 找番、 追更、看最新一集、拿最新一季第一集、帮我下载、只要磁力、先找资源别下载、或者查看刚才下载怎么样; also use it for m... It is an AI Agent Skill for Claude Code / OpenClaw, with 100 downloads so far.

How do I install Anime Finder?

Run "/install anime-finder" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Anime Finder free?

Yes, Anime Finder is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Anime Finder support?

Anime Finder is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Anime Finder?

It is built and maintained by TiggyChan (@tiggy-chan); the current version is v1.4.0.

💬 Comments