Resource Hunter

Find the best public download route for movies, TV, anime, music, software, books, pan links, magnets, and public video URLs. Uses high-recall routing, quali...

Security Scan

The skill's code, instructions, and requirements are internally consistent with its stated purpose (discovering public download routes and probing public video URLs); it will perform network queries and write caches/downloads to disk, which is expected but has operational and legal implications the user should consider.

Purpose & Capability

Name/description match the delivered artifacts: CLI entrypoint, source adapters, ranking, JSON output, and video pipeline. No unrelated credentials, binaries, or config paths are requested. The included source list (pan/torrent/video sites) aligns with the skill's goal.

Instruction Scope

SKILL.md and CLI instruct the agent to query public sources, probe/inspect public video URLs, and optionally download videos. The runtime rules explicitly avoid logins, cookies, or APIs and require yt-dlp/ffmpeg for downloads. The skill will perform network requests to many public endpoints and will cache results and video manifests locally; this behavior is within scope but worth highlighting because it involves external network activity and file writes.

Install Mechanism

No install spec is present (instruction-only skill as far as packaging), but the repository includes the actual Python scripts which will be executed from the skill directory. Nothing in the manifest downloads or executes remote installers automatically. This is proportionate to the skill's function.

Credentials

The skill requests no credentials or special env vars. It optionally uses OPENCLAW_WORKSPACE if present to determine storage location and checks for yt-dlp/ffmpeg binaries before downloads. No access to unrelated secrets or system-level credentials is requested.

Persistence & Privilege

always:false (normal). The skill creates and uses an SQLite cache and stores video manifests and optionally downloads files into a storage root under the workspace (~/.openclaw/workspace by default). This is expected but means the skill will write to disk and can consume storage and network bandwidth.

Scan Findings in Context

[pre-scan-injection-signals] expected: Static pre-scan reported no injection signals. That's consistent with the repository being a visible Python implementation that performs HTTP queries and local caching.

Assessment

This skill appears to be what it claims: a multi-source public-resource searcher that probes public video URLs and can download when yt-dlp and ffmpeg are available. Before installing, consider: 1) Legal/ethical risk — it can surface and help download copyrighted material; do not use it to bypass access controls. 2) Network exposure — it will query many public sites (including third-party aggregators) which can leak the queries you run; treat those queries as network traffic that may reveal what you searched for. 3) Disk writes — it creates an SQLite cache and may write downloaded video artifacts under your OpenClaw workspace (OPENCLAW_WORKSPACE or ~/.openclaw/workspace); review/clean that directory if needed. 4) Hostile content risk — results may link to malicious files; avoid automatic downloads on a production host and run downloads in a sandboxed environment. 5) Binaries — yt-dlp and ffmpeg are optional but required for downloading/subtitles; the skill checks for them before attempting downloads. If you accept those operational and legal considerations, the skill's requirements and behavior are coherent. If you want tighter safety, run it in an isolated environment with limited network and filesystem access.

Public resource discovery for agents.
Use it to find the best public download route, not to bypass access controls.

Use this skill when the user wants to

• Find public pan links, torrent results, or magnets
• Compare releases for quality, confidence, and source health
• Search movies, TV, anime, music, software, or books
• Inspect a public video URL and optionally download it if the machine is ready
• Get compact chat output or structured JSON for downstream tools

Do not use this skill for

• Login-only sites, private accounts, cookies, invite-only trackers, or private libraries
• DRM, captchas, bypassing access controls, or restricted content
• Guarantees about legality, permanence, or long-term availability

What this skill gives the agent

• High-recall multi-source search across pan and torrent channels
• Tiered ranking: top, related, and suppressed risky recall
• Quality-aware sorting with title-family matching, season/episode checks, and release parsing
• Live source health, degraded-source penalties, and circuit breaking
• JSON v3 output with canonical_identity, evidence, source_health, and upstream_source

Main entrypoint

SKILL_DIR="$(openclaw skills path resource-hunter)/scripts"
python3 "$SKILL_DIR/hunt.py" search "<query>"

Quick start

python3 "$SKILL_DIR/hunt.py" search "Oppenheimer 2023" --4k
python3 "$SKILL_DIR/hunt.py" search "Breaking Bad S01E01" --tv
python3 "$SKILL_DIR/hunt.py" search "Attack on Titan" --anime --sub
python3 "$SKILL_DIR/hunt.py" search "Jay Chou Fantasy lossless" --music
python3 "$SKILL_DIR/hunt.py" search "Adobe Photoshop 2024" --software --channel pan
python3 "$SKILL_DIR/hunt.py" video probe "https://www.bilibili.com/video/BV..."
python3 "$SKILL_DIR/hunt.py" benchmark

Default routing

• Movie: pan first, torrent second
• TV: EZTV/TPB first, then pan supplements
• Anime: Nyaa first, then pan supplements
• Music, software, book: pan first, torrent fallback
• Public video URL: skip search and route directly to video probe / video info

Output contract

• Default text: only top and related results are shown
• Risky recall stays suppressed in text and is still available in JSON
• --json: JSON v3 with schema_version, query, intent, plan, results, suppressed, warnings, source_status, and meta
• --json-version 2: temporary compatibility mode for older consumers

Runtime requirements

• Search, ranking, and source diagnostics: no API keys required
• Public video inspect/probe: yt-dlp optional but recommended
• Public video download/subtitle: requires both yt-dlp and ffmpeg

Agent rules

• Prefer the main hunt.py entrypoint over lower-level wrappers
• Use --quick when the user wants a short answer
• Use --json when another tool or script will consume the output
• If the input is a public video URL, do not search pan/torrent first; go straight to the video pipeline
• If the user asks to actually download a public video URL, first verify yt-dlp and ffmpeg are installed; otherwise return the route plus the missing dependency requirement
• If the user wants only pan or only torrent, set --channel pan or --channel torrent
• Prefer benchmark-backed changes over heuristic tweaks without regression coverage

Legacy compatibility

Legacy wrappers still exist for one compatibility cycle:

python3 "$SKILL_DIR/pansou.py" "<query>"
python3 "$SKILL_DIR/torrent.py" "<query>"
python3 "$SKILL_DIR/video.py" info "<url>"

References

• Detailed usage: references/usage.md
• Internal architecture and JSON schema: references/architecture.md
• Source coverage and routing notes: references/sources.md