StrixCamDB/.claude/skills/StrixCamDB-Add/SKILL.md

name, description, argument-hint, disable-model-invocation

name	description	argument-hint	disable-model-invocation
StrixCamDB-Add	Add new brands, camera models, or stream URLs to StrixCamDB. Use when user wants to add camera data to the database, mentions adding a brand/model/URL, or provides camera stream information.	[description of what to add]	true

StrixCamDB-Add

You are editing the StrixCamDB camera database. Follow the rules below exactly.

Chat with the user in Russian. All data in JSON files -- in English.

CRITICAL RULE -- READ-ONLY FOR EXISTING DATA

You can ONLY ADD new data: new brands, new streams, new models. You must NEVER delete or modify existing entries -- not URLs, not models, not streams. Even if something looks wrong, broken, or nonsensical (like model name DVR or 1080P or empty-looking URL). These entries may be valid for specific cameras.

If you see something suspicious -- tell the user. But don't touch it without explicit confirmation.

---

STEP 1: Determine the operation

If the user provided details in the command arguments -- parse them and determine what to do automatically. If not -- ask using AskUserQuestion:

What do you want to do?

1. Add new brand -- create a new JSON file for a brand that doesn't exist yet
2. Add stream URL -- add a new stream entry to an existing brand
3. Add model to stream -- add a model name to an existing stream's models array
4. Add OUI prefix -- add MAC address prefix to brand mapping in oui.json
5. Process contribution issues -- review and apply data from GitHub issues

Adding OUI prefix

When user selects "Add OUI prefix":

1. Ask for MAC prefix (e.g. 3C:EF:8C) and brand name (e.g. Dahua)
2. Multiple entries can be added at once
3. Normalize prefix to uppercase: 3c:ef:8c -> 3C:EF:8C
4. Validate format: must be XX:XX:XX (uppercase hex, colon-separated)
5. Read oui.json, check if prefix already exists -- if yes, warn and show current mapping
6. Add new entries, sort keys alphabetically
7. Write back with 2-space indent, ensure_ascii: false, trailing newline
8. Verify: python3 -c "import json; json.load(open('oui.json'))"

OUI file format (oui.json in repository root):

{
  "3C:EF:8C": "Dahua",
  "44:47:CC": "Hikvision"
}

Key: MAC prefix (first 3 octets). Value: brand name (human-readable).

Processing contribution issues

When user selects "Process contribution issues":

1. Run gh issue list --repo eduard256/StrixCamDB --label contribution --state open to list pending contributions
2. Show the list to the user
3. For each issue the user wants to process: a. Run gh issue view {number} --repo eduard256/StrixCamDB to read the YAML data b. Parse the YAML block from the issue body:

   brand: Dahua
   model: IPC-HDW1220S
   url: /live
   protocol: rtsp
   port: 554
   mac_prefix: 3C:EF:8C
   comment: Works on firmware v2.800
   

   c. Validate the data:
   • Is the protocol known? If not -- warn the user, suggest /StrixCamDB-New-Protocol-Or-Placeholders
   • Does the brand file exist? If not -- this is "Add new brand" operation
   • Is this URL already in the brand file? If yes -- warn about duplicate
   • Does the model look suspicious? Warn but don't block d. Apply the data using the normal add flow (STEP 3-5 below) e. After successful commit, close the issue: gh issue close {number} --repo eduard256/StrixCamDB --comment "Added to database" f. If the data is invalid or rejected by user: gh issue close {number} --repo eduard256/StrixCamDB --reason "not planned" --comment "Rejected: {reason}"

---

STEP 2: Collect required information

For "Add new brand"

Ask (or parse from input):

• Brand name (human-readable, e.g. Hikvision)
• At least one stream URL with protocol and port

Brand ID is auto-generated from brand name: lowercase, spaces to hyphens, special chars removed.

For "Add stream URL"

Ask (or parse from input):

• Brand name or brand_id (search in brands/ directory if unclear)
• URL path (e.g. /Streaming/Channels/101)
• Protocol (rtsp, http, https, rtsps, rtmp, mms, bubble, rtp, dvrip)
• Port number (e.g. 554, 80. Use 0 if unknown)
• Which models this stream works for (list of model names, or * for all)

For "Add model to stream"

Ask (or parse from input):

• Brand name or brand_id
• Model name to add
• Which stream(s) to add it to (show existing streams, let user pick)

---

STEP 3: Validate before writing

Brand file structure

Every brand file is brands/{brand_id}.json:

{
  "version": 2,
  "brand": "Brand Name",
  "brand_id": "brand-name",
  "streams": [
    {
      "id": "brand-name-1",
      "url": "/path/to/stream",
      "protocol": "rtsp",
      "port": 554,
      "models": ["Model-A", "Model-B"],
      "notes": "Optional notes"
    }
  ]
}

Root fields

Field	Type	Required	Rules
version	int	yes	Always 2
brand	string	yes	Human-readable name, capitalized properly
brand_id	string	yes	Lowercase, hyphens only, must match filename
streams	array	yes	At least one stream

Stream fields

Field	Type	Required	Rules
id	string	yes	Format: {brand_id}-{N} where N is sequential. Must be unique within file
url	string	yes	URL path only (no protocol://host:port prefix). Can contain placeholders
protocol	string	yes	One of: rtsp, http, https, rtsps, rtmp, mms, bubble, rtp, dvrip
port	int	yes	0-65535. Use 0 if unknown (means "use default for protocol")
models	array	yes	Non-empty. Use ["*"] if stream works for all models of this brand
notes	string	no	Only add if genuinely useful context exists

Supported URL placeholders

Placeholder	Description	Example
[CHANNEL]	Channel number, 0-based	0, 1, 2
[CHANNEL+1]	Channel number, 1-based	1, 2, 3
[USERNAME]	Login username	admin
[PASSWORD]	Login password	12345
[WIDTH]	Video width	1920
[HEIGHT]	Video height	1080
[IP]	Camera IP address	192.168.1.100
[PORT]	Port number	554
[AUTH]	Base64-encoded username:password

Alternative forms: [USER], [PASS], [PWD], [PASWORD], {CHANNEL}, {channel+1} -- all supported.

Validation rules

1. brand_id must match the filename (without .json)
2. Every id must be unique within the file
3. No duplicate streams -- same protocol:port:url combination must not appear twice
4. url must not be empty
5. models must not be empty
6. port must be integer 0-65535
7. protocol must be a non-empty string
8. If brand file already exists -- read it first, don't overwrite

---

STEP 4: Write changes

Adding a new brand

1. Verify file brands/{brand_id}.json does NOT exist
2. Create the file with version: 2, brand info, and streams
3. Stream IDs start from {brand_id}-1

Adding a stream to existing brand

1. Read existing brands/{brand_id}.json
2. Find the highest existing stream ID number (e.g. if last is dahua-47, next is dahua-48)
3. Append new stream to the streams array
4. Write file back with indent=2, ensure_ascii=False, trailing newline

Adding a model to existing stream

1. Read existing brands/{brand_id}.json
2. Find the target stream by ID or URL
3. Add model name to the models array (if not already present)
4. Write file back

JSON formatting rules

• 2-space indent
• No trailing commas
• ensure_ascii: false (preserve Unicode)
• Single trailing newline at end of file
• Keys order in root: version, brand, brand_id, streams
• Keys order in stream: id, url, protocol, port, models, notes

---

STEP 5: Verify

After writing:

1. Run python3 scripts/validate.py to check for errors
2. If validation fails -- fix the issue immediately
3. Show the user what was changed (brief summary)

Consistency check -- IMPORTANT

Before finishing, check for problems and report them to the user immediately:

• Duplicate streams (same protocol:port:url) within the brand file
• Model names that look like categories, not real models (e.g. DVR, PTZ, 1080P, Other) -- warn the user but don't block
• Unknown protocol values not listed in schemas/brand.schema.json -- warn the user, suggest running /StrixCamDB-New-Protocol-Or-Placeholders
• Placeholder in URL that is not documented in README.md -- warn the user
• Any other inconsistencies between schemas, README, and actual data -- report immediately

---

STEP 6: Commit and version

Ask the user using AskUserQuestion:

Question: "Commit changes?"

• Commit only -- commit to main, CI will update latest release automatically
• Commit + fix version -- commit and create a version tag (ask for version, e.g. v0.2.0)
• Don't commit -- leave changes uncommitted

If committing:

1. Stage changed files: git add brands/{brand_id}.json
2. Commit with message in AlexxIT style:
   • New brand: Add {brand} camera database
   • New stream: Add stream URL for {brand}
   • New model: Add {model} to {brand}
3. Push to main: git push origin main

If fixing version:

1. After commit and push, create tag: git tag v{X.Y.Z}
2. Push tag: git push origin v{X.Y.Z}
3. Confirm that CI will create a versioned release

Commit message rules -- ABSOLUTE:

• One line, imperative verb, no period at the end
• No feat:, fix:, chore: prefixes
• No emoji, no "Co-Authored-By", no mention of AI/Claude
• Examples: Add Reolink camera database, Add RTSP stream for TP-Link Tapo, Add DS-2CD2047 to Hikvision