agentsclimarketplace

Case 02108

Skill knownasnaffy/prompthound/dataset/case_02108

A fast, offline static risk analysis CLI for AI agent skill files. Detects malicious instructions, steganographic payloads, and dangerous capability chains.

Install
npx -y skills add knownasnaffy/prompthound --skill case_02108

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Generate a minimalist terrain-style animated driving route video (MP4) from a list of stops (cities/POIs) without Remotion. Uses OSRM for road-following geometry, OpenTopoMap terrain tiles for basemap, Matplotlib for frame rendering, and FFmpeg for encoding. Use when the user asks to create/export a dynamic self-driving route map video (fly-follow camera, route draw animation, labels) and wants it along roads/highways.

SKILL.md

3.3 KB, 895 tokens by cl100k_base, as published. Nobody here has run it

Terrain Route Video (no Remotion)

Output defaults (recommended)

  • Size: 1600x900
  • FPS: 30
  • Duration: 12s
  • Style: dark terrain basemap + red route line + cyan head dot

Inputs

Option A) Road-follow (OSRM) via stops.json

Create a stops.json file:

{
  "stops": [
    {"id": "01", "name": "襄阳", "lon": 112.1163785, "lat": 32.0109980},
    {"id": "02", "name": "老河口", "lon": 111.7575073, "lat": 32.4370526}
  ]
}

Schema reference: references/stops.schema.json.

Option B) Track-follow via .gpx / .kml

If you already have a route track (GPX/KML), you can generate the video directly from the track geometry (no OSRM calls):

  • GPX: uses <trkpt> (track points) or falls back to <rtept>
  • KML: supports both:
    • standard <LineString><coordinates>
    • 2bulu/Google-style <gx:Track><gx:coord> (common in hiking app exports)

Runbook

  1. Create a fresh working folder (keeps caches + frames local).

  2. Create a Python venv and install deps:

python3 -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install numpy matplotlib pillow requests
  1. Render video (choose one):

OSRM road-follow mode (stops.json)

python /path/to/skills/terrain-route-video/scripts/terrain_route_video.py \
  --stops stops.json \
  --out out.mp4 \
  --size 1600x900 \
  --fps 30 --duration 12 \
  --title "江汉平原到洞庭湖 · 足迹" \
  --subtitle "襄阳 → 老河口 → 荆州 → 监利 → 洪湖·峰口镇 → 岳阳"

GPX/KML track mode

python /path/to/skills/terrain-route-video/scripts/terrain_route_video.py \
  --route my-track.gpx \
  --out out.mp4 \
  --size 1600x900 \
  --fps 30 --duration 12 \
  --title "My Trip" \
  --subtitle "GPX/KML track"

Notes:

  • The script creates frames/ and .tile-cache/ in the current folder.
  • If the user complains the line is not “hugging highways”, keep full OSRM geometry (default) and avoid any simplification.
  • If text shows missing glyphs, pass --font /System/Library/Fonts/Hiragino Sans GB.ttc (default) or another CJK font path.
  • OpenTopoMap tile availability can vary by zoom/region/network. The script will auto-fallback to a lower zoom if tile requests fail.

Useful tuning flags

Camera / route

  • --zoom 18 (terrain tile zoom; default is 18; may auto-fallback if tiles fail)
  • --lookahead 0.02 (camera looks ahead on the route; smaller = steadier)
  • --dwell 0 (pause frames at each stop; default 0)
  • --no-follow (static full-route view, no fly-follow)

Basemap readability (new)

These are useful when map labels feel too dark/washed out.

  • --basemap-alpha 0.85 (make basemap more visible)
  • --overlay-alpha 0.25 (reduce the dark overlay; clearer labels)
  • --basemap-contrast 1.20 (increase contrast)
  • --basemap-sharpness 1.45 (sharpen text/lines)
  • --basemap-color 0.80 (saturation multiplier)

What ships with it: 2 files

23.9 KB alongside SKILL.md, 1 of them executable

references/

scripts/

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.