Skip to content
Skillv1.0.0

remove-ai-watermarks

Provenance-hygiene inspects and strips AI provenance marks from named files the operator owns on this phone: invisible Unicode (Layer A) in text, and EXIF/XMP/C2PA-style tags via ffmpeg/exiftool on a

by CynicalTyr(0) 0 installs
Free
Sign in to install

Free account. Installing gives you the manifest plus copy-paste snippets.

See reviews

About

Imported from CynicalTyr/termux-cursor-integration (skills/remove-ai-watermarks/SKILL.md). Install upstream with npx skills add CynicalTyr/termux-cursor-integration --skill remove-ai-watermarks. Copyright stays with the author.

Remove AI watermarks (this phone)

Provenance hygiene for content the operator owns or is authorized to process. Inspect first, then clean. Do not claim output is undetectable or human-written.

Read when needed:

SCRIPTS="${HOME}/.cursor/skills/remove-ai-watermarks/scripts"
python3 "$SCRIPTS/inspect_text.py" path.txt
python3 "$SCRIPTS/clean_text.py" path.txt -o path.cleaned.txt

If skills were copied but $HOME is not /root, use the absolute scripts path.

Pre-flight

  • Named files only. Do not audit all of /storage/emulated/0/DCIM or Pictures.
  • Prefer sidecars (*.cleaned.*). In-place only when asked.
  • Pixel-domain GPU inpainters and untrusted pip “unwatermark” clones are out of scope. Done when: ownership is stated and the refuse table applied.

Refuse / still-clean

User intent Action
Own drafts: strip invisible Unicode, EXIF/XMP, container tags Proceed
Academic fraud / “make this look human-written” Refuse the deception
Walk the whole camera roll Refuse; ask for names
Crawl a website to strip marks Refuse

Workflow

  1. Classify: text vs image vs video (named path). Done when: kind is stated.
  2. Text: inspect then clean with the scripts above. Re-inspect. Done when: inspect report shows counts and the sidecar exists.
  3. Image/video: exiftool -s FILE and/or ffprobe. Strip tags with media-edit sidecar commands (-map_metadata -1, exiftool -all=). Visible delogo only on a short named clip. Done when: sidecar exists; report separates verifiable vs best-effort.
  4. Honesty: a stripped C2PA/EXIF block does not mean “no AI left.” Done when: the report uses those words if metadata was removed.

Related skills

  • media-edit ffmpeg/exiftool mechanics
  • file-edit text files under Termux home or /root
  • android-storage do not scan entire gallery trees

Use it

Copy one of these into your project. Installing also returns the manifest and these snippets.

yaml
targets:
  - https://api.opensmartroute.ai/api/v1/registry/cynicaltyr-termux-cursor-integration-remove-ai-watermarks/manifest   # or paste the manifest below

Manifest

An Open Capability Manifest: the router reads it to know what this does, what it costs and when to pick it.

cynicaltyr-termux-cursor-integration-remove-ai-watermarks.ocm.jsonjson
{
  "ocm": "1",
  "id": "cynicaltyr-termux-cursor-integration-remove-ai-watermarks",
  "kind": "skill",
  "name": "remove-ai-watermarks",
  "description": "Provenance-hygiene inspects and strips AI provenance marks from named files the operator owns on this phone: invisible Unicode (Layer A) in text, and EXIF/XMP/C2PA-style tags via ffmpeg/exiftool on a named photo or short clip. Use when the user asks to strip watermarks, remove Content Credentials, or clean AI metadata on a named path. Do not walk all of DCIM.",
  "publisher": "CynicalTyr",
  "version": "1.0.0",
  "capabilities": {
    "domains": [
      "general"
    ],
    "tags": [
      "skill-md",
      "android",
      "watermark",
      "unicode",
      "c2pa",
      "github"
    ],
    "languages": [
      "en"
    ]
  },
  "quality_prior": 0.6,
  "examples": [
    "Provenance-hygiene inspects and strips AI provenance marks from named files the operator owns on this phone: invisible Unicode (Layer A) in text, and EXIF/XMP/C2PA-style tags via ffmpeg/exiftool on a named photo or short clip. Use when the user asks to strip watermarks, remove Content Credentials, or clean AI metadata on a named path. Do not walk all of DCIM."
  ],
  "primary": false,
  "metadata": {
    "source": {
      "provider": "github",
      "repository": "https://github.com/CynicalTyr/termux-cursor-integration",
      "path": "skills/remove-ai-watermarks/SKILL.md",
      "ref": "e08a174b0bad278559efd50cc23444a304969887",
      "url": "https://github.com/CynicalTyr/termux-cursor-integration/blob/e08a174b0bad278559efd50cc23444a304969887/skills/remove-ai-watermarks/SKILL.md",
      "key": "CynicalTyr/termux-cursor-integration/skills/remove-ai-watermarks/SKILL.md"
    },
    "compatibility": "Ubuntu in F-Droid Termux; python3, ffmpeg, optionally exiftool."
  },
  "instructions": "# Remove AI watermarks (this phone)\n\nProvenance hygiene for content the operator **owns or is authorized to process**.\nInspect first, then clean. Do not claim output is undetectable or human-written.\n\nRead when needed:\n\n- [references/ethics.md](references/ethics.md)\n- [references/mark-classes.md](references/mark-classes.md)\n\n```bash\nSCRIPTS=\"${HOME}/.cursor/skills/remove-ai-watermarks/scripts\"\npython3 \"$SCRIPTS/inspect_text.py\" path.txt\npython3 \"$SCRIPTS/clean_text.py\" path.txt -o path.cleaned.txt\n```\n\nIf skills were copied but `$HOME` is not `/root`, use the absolute scripts path.\n\n## Pre-fli",
  "cost": {
    "context_tokens": 504
  }
}

Fetch it by URL: GET /api/v1/registry/cynicaltyr-termux-cursor-integration-remove-ai-watermarks/manifest?version=1.0.0

Reviews

Star ratings from people who tried it. One review per account; edit yours any time.

No reviews yet. Install it, try it, and be the first to rate it.