[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77VPF50EE5246YN4DN9WK1K":3,"skill-same:01M77VPF50EE5246YN4DN9WK1K":59},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":36,"commitSha":37,"skillDirectory":38,"forksCount":39,"repositoryHtmlUrl":40,"repositoryHomepageUrl":41,"ownerHtmlUrl":42,"ownerBlogUrl":6,"ownerDescription":43,"updatedAt":44,"install":45,"resources":58},"01M77VPF50EE5246YN4DN9WK1K","subscription-killer",null,"Analyse a bank transactions CSV to detect recurring subscriptions, score cancellation priority, and surface actionable savings recommendations. Designed as the first module in a broader personal-finance agent; bank data parsing is deliberately decoupled so future savings \u002F deposit \u002F investment skills can reuse it.","Lord1Egypt\u002Fawesome-skill-forge","Lord1Egypt","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F99746359?v=4","01M7592YDNT0XK9WJ0T2MQVPDN","Mohamed Mounir Ali AL Keshita","01M75GRFNHC8VZNWEB8DKDE433",3,"MIT","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fs\u002Fsubscription-killer\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: subscription-killer\ndescription: >\n  Analyse a bank transactions CSV to detect recurring subscriptions, score\n  cancellation priority, and surface actionable savings recommendations.\n  Designed as the first module in a broader personal-finance agent; bank data\n  parsing is deliberately decoupled so future savings \u002F deposit \u002F investment\n  skills can reuse it.\nversion: 1.1.0\nmetadata:\n  openclaw:\n    requires:\n      env: []\n      bins:\n        - python3\n    envVars:\n      - name: SUBSCRIPTION_KILLER_CURRENCY\n        required: false\n        description: >\n          ISO 4217 currency code for display (default: GBP). E.g. USD, EUR, GBP.\n    primaryEnv: SUBSCRIPTION_KILLER_CURRENCY\n---\n\n# Subscription Killer\n\nAnalyse a bank transactions CSV file and surface every recurring subscription,\nranked by cancellation \u002F downgrade priority with estimated monthly savings.\n\n## Skill overview\n\nThis skill takes a single CSV file of bank transactions and:\n\n1. Sniffs the CSV structure to detect column names automatically.\n2. Normalises merchant names using a curated alias dictionary plus fuzzy\n   matching (Levenshtein distance ≤ 2 tokens).\n3. Detects recurring charges by cadence (monthly \u002F quarterly \u002F annual) and\n   amount consistency (±2 % tolerance).\n4. Scores each subscription by confidence (0–100).\n5. Ranks subscriptions by a composite priority score: spend × inactivity risk ×\n   price-creep factor × duplication penalty.\n6. Outputs a structured report with clear cancellation and downgrade actions.\n\n---\n\n## Input format\n\nThe skill accepts any CSV export from a personal bank account. Common formats\nare auto-detected:\n\n| Bank \u002F Provider | Typical columns |\n|-----------------|-----------------|\n| Monzo | Date, Name, Amount, Category |\n| Starling | Date, Counter Party, Amount, Balance |\n| Revolut | Started Date, Description, Amount, Currency |\n| Standard OFX export | Date, Description, Debit Amount |\n| Generic | date, merchant \u002F description, amount (minimum required) |\n\nColumn detection priority:\n- **Date**: looks for columns named `date`, `transaction date`, `started date`,\n  `posted date` (case-insensitive). Falls back to first column that parses as\n  ISO 8601 or DD\u002FMM\u002FYYYY.\n- **Amount**: looks for `amount`, `debit amount`, `debit`. Ignores credit \u002F\n  top-up rows (positive amounts in debit-positive conventions).\n- **Merchant**: looks for `name`, `description`, `merchant`, `counter party`,\n  `payee`. Falls back to longest non-numeric column.\n\n---\n\n## Core logic\n\n### Step 1 — Parse CSV\n\n```python\nimport csv, re, difflib\nfrom datetime import datetime, timedelta\nfrom collections import defaultdict\n\ndef sniff_columns(header: list[str]) -> dict:\n    \"\"\"Return {'date': col, 'amount': col, 'merchant': col} best guesses.\"\"\"\n    header_lower = [h.lower().strip() for h in header]\n    DATE_HINTS    = ['date', 'transaction date', 'started date', 'posted date']\n    AMOUNT_HINTS  = ['amount', 'debit amount', 'debit']\n    MERCHANT_HINTS = ['name', 'description', 'merchant', 'counter party',\n                      'payee', 'narrative']\n    def first_match(hints):\n        for hint in hints:\n            if hint in header_lower:\n                return header[header_lower.index(hint)]\n        return None\n    return {\n        'date':     first_match(DATE_HINTS)     or header[0],\n        'amount':   first_match(AMOUNT_HINTS)   or header[1],\n        'merchant': first_match(MERCHANT_HINTS) or header[2],\n    }\n\ndef parse_amount(raw: str) -> float | None:\n    \"\"\"Return a positive debit amount, or None for credits \u002F errors.\"\"\"\n    cleaned = re.sub(r'[£$€,\\s]', '', raw.strip())\n    if not cleaned:\n        return None\n    try:\n        val = float(cleaned)\n        # Debit-positive CSV (most UK banks): only keep debits (negatives after\n        # sign flip are credits — skip them)\n        # Some CSVs use negative for debits; normalise to positive debit.\n        return abs(val) if val != 0 else None\n    except ValueError:\n        return None\n```\n\n### Step 2 — Normalise merchant names\n\nUse a two-pass approach:\n\n**Pass 1 — Known alias dictionary** (curated, highest-confidence)\n\n```python\nMERCHANT_ALIASES = {\n    # Streaming\n    r'netflix':                 'Netflix',\n    r'spotify':                 'Spotify',\n    r'apple\\.com\u002Fbill':         'Apple Subscriptions',\n    r'itunes':                  'Apple Subscriptions',\n    r'amazon prime':            'Amazon Prime',\n    r'amzn\\s?prime':            'Amazon Prime',\n    r'disney\\+|disneyplus':     'Disney+',\n    r'youtube premium':         'YouTube Premium',\n    r'hbo|max\\.com':            'Max (HBO)',\n    r'paramount':               'Paramount+',\n    r'dazn':                    'DAZN',\n    # Music\n    r'tidal':                   'Tidal',\n    r'deezer':                  'Deezer',\n    r'soundcloud':              'SoundCloud',\n    # Productivity \u002F SaaS\n    r'github':                  'GitHub',\n    r'notion':                  'Notion',\n    r'slack':                   'Slack',\n    r'dropbox':                 'Dropbox',\n    r'google\\s?(one|storage|workspace)': 'Google One\u002FWorkspace',\n    r'microsoft 365|office 365|msft': 'Microsoft 365',\n    r'adobe':                   'Adobe Creative Cloud',\n    r'figma':                   'Figma',\n    r'zoom':                    'Zoom',\n    r'1password|lastpass|bitwarden': 'Password Manager',\n    r'nordvpn|expressvpn|surfshark':  'VPN Service',\n    # News \u002F Reading\n    r'medium':                  'Medium',\n    r'substack':                'Substack',\n    r'kindle unlimited':        'Kindle Unlimited',\n    r'audible':                 'Audible',\n    r'the times|thetimes':      'The Times',\n    r'financial times|ft\\.com': 'Financial Times',\n    r'economist':               'The Economist',\n    # Fitness\n    r'peloton':                 'Peloton',\n    r'strava':                  'Strava',\n    r'myfitnesspal':            'MyFitnessPal',\n    r'calm|headspace':          'Meditation App',\n    # Utilities \u002F services\n    r'amazon web services|aws': 'AWS',\n    r'google cloud|gcp':        'Google Cloud',\n    r'digitalocean':            'DigitalOcean',\n    r'cloudflare':              'Cloudflare',\n    r'railway\\.app':            'Railway',\n    r'openai':                  'OpenAI',\n    r'anthropic':               'Anthropic',\n}\n\ndef normalize_merchant(raw: str) -> str:\n    \"\"\"Return a clean, canonical merchant name.\"\"\"\n    s = raw.lower().strip()\n    for pattern, canonical in MERCHANT_ALIASES.items():\n        if re.search(pattern, s, re.IGNORECASE):\n            return canonical\n    # Pass 2 — strip noise tokens (card numbers, references, country codes)\n    cleaned = re.sub(\n        r'\\b(www|com|co\\.uk|ltd|limited|uk|us|gb|inc|gmbh)\\b'\n        r'|\\d{4,}'      # card \u002F ref numbers\n        r'|\\*+\\S*'      # asterisk-prefixed tokens\n        r'|[^a-z0-9 ]', # special chars\n        '', s\n    ).strip().title()\n    return cleaned or raw.title()\n```\n\n**Pass 2 — Fuzzy deduplication** across the transaction list\n\nAfter normalisation, cluster remaining merchant names by token-set similarity\n(`difflib.SequenceMatcher` ratio > 0.82). The most-frequent variant in each\ncluster becomes the canonical name.\n\n### Step 3 — Detect recurring charges\n\n```python\ndef detect_cadence(dates: list[datetime]) -> str | None:\n    \"\"\"\n    Return 'monthly', 'quarterly', 'annual', or None.\n    Requires ≥ 2 transactions for monthly\u002Fquarterly, ≥ 1 for annual hinting.\n    \"\"\"\n    if len(dates) \u003C 2:\n        # Single occurrence — flag as potential annual if amount > 30\n        return 'annual_candidate'\n    gaps = sorted([(dates[i+1] - dates[i]).days for i in range(len(dates)-1)])\n    median_gap = gaps[len(gaps)\u002F\u002F2]\n    if 25 \u003C= median_gap \u003C= 35:   return 'monthly'\n    if 85 \u003C= median_gap \u003C= 100:  return 'quarterly'\n    if 340 \u003C= median_gap \u003C= 380: return 'annual'\n    return None\n\ndef amounts_consistent(amounts: list[float], tolerance: float = 0.02) -> bool:\n    \"\"\"True if all amounts are within ±2 % of the median.\"\"\"\n    if not amounts:\n        return False\n    median = sorted(amounts)[len(amounts)\u002F\u002F2]\n    return all(abs(a - median) \u002F median \u003C= tolerance for a in amounts)\n```\n\n### Step 4 — Confidence scoring (0–100)\n\n| Factor | Max points | Logic |\n|--------|-----------|-------|\n| Cadence regularity | 40 | All gaps within ±3 days of median |\n| Amount consistency | 30 | All amounts within ±2 % |\n| Known merchant match | 20 | Exact alias dictionary hit |\n| Transaction count | 10 | log₂(count) × 3, capped at 10 |\n\nConfidence ≥ 70 → **Confirmed subscription**\nConfidence 40–69 → **Probable subscription**\nConfidence \u003C 40 → **Possible \u002F flag for review**\n\n### Step 5 — Priority ranking\n\nEach confirmed\u002Fprobable subscription gets a composite priority score:\n\n```\npriority = monthly_cost\n         × price_creep_multiplier   # max(amounts) \u002F min(amounts); 1.0 = stable\n         × duplication_penalty       # 1.5 if same category has 2+ subscriptions\n         × inactivity_risk           # 1.3 if last charge > 45 days ago\n```\n\nHigher priority score = stronger cancellation candidate.\n\n**Special flags:**\n- **Trial trap**: single charge in the last 7 days from a previously-unseen\n  merchant, followed by a recurring pattern. Flag as \"check for auto-renew\".\n- **Price creep**: amount has increased > 5 % over the observation window.\n  Flag with the exact amount delta.\n- **Annual renewal due**: annual-cadence subscription whose last charge was\n  300–365 days ago. Flag with estimated renewal date.\n- **Unknown merchant**: low-confidence recurring charge with no alias match.\n  Flag for manual review.\n\n---\n\n## Output schema\n\nThe skill returns a structured report. Print to stdout as JSON for downstream\nagent consumption, or render the human-readable summary below.\n\n```json\n{\n  \"summary\": {\n    \"observation_window_days\": 90,\n    \"total_subscriptions\": 12,\n    \"confirmed\": 9,\n    \"probable\": 2,\n    \"flagged_for_review\": 1,\n    \"monthly_spend_gbp\": 187.43,\n    \"annual_spend_gbp\": 2249.16,\n    \"potential_monthly_saving_gbp\": 54.20\n  },\n  \"subscriptions\": [\n    {\n      \"merchant\": \"Adobe Creative Cloud\",\n      \"cadence\": \"monthly\",\n      \"median_amount\": 54.99,\n      \"last_charge\": \"2025-04-01\",\n      \"confidence\": 95,\n      \"priority_score\": 89.2,\n      \"flags\": [\"price_creep\"],\n      \"price_creep_delta\": 5.00,\n      \"cancel_url\": \"https:\u002F\u002Faccount.adobe.com\u002Fplans\",\n      \"actions\": [\"cancel\", \"downgrade\"]\n    }\n  ],\n  \"top_cancellation_targets\": [\"Adobe Creative Cloud\", \"...\"],\n  \"unknown_merchants\": [\"TRSF 48291 REF77\"],\n  \"annual_renewals_due\": []\n}\n```\n\n### Human-readable summary (terminal \u002F Telegram)\n\n```\n╔══════════════════════════════════════════╗\n║        SUBSCRIPTION KILLER  v1.0         ║\n╠══════════════════════════════════════════╣\n║  12 subscriptions  •  £187\u002Fmo  •  £2,249\u002Fyr\n╚══════════════════════════════════════════╝\n\nTOP CANCELLATION TARGETS\n1. Adobe Creative Cloud   £54.99\u002Fmo  ↑ price crept +£5.00\n2. Calm                   £39.99\u002Fyr  last used 3 months ago\n3. LinkedIn Premium       £29.99\u002Fmo  duplicate: 2 job-search tools\n\nPOTENTIAL SAVING:  £54\u002Fmo  •  £648\u002Fyr\n\nUNKNOWN MERCHANTS (review manually)\n• TRSF 48291 REF77  — £12.99\u002Fmo  (confidence 38%)\n```\n\n---\n\n## Cancel guidance (top 50 services)\n\nThe skill bundles a `cancel_urls.json` file mapping canonical merchant names\nto their self-serve cancellation pages. This removes friction for the user.\n\n```json\n{\n  \"Netflix\":                \"https:\u002F\u002Fwww.netflix.com\u002Fcancelplan\",\n  \"Spotify\":                \"https:\u002F\u002Fwww.spotify.com\u002Faccount\u002Fsubscription\u002Fcancel\",\n  \"Adobe Creative Cloud\":   \"https:\u002F\u002Faccount.adobe.com\u002Fplans\",\n  \"Amazon Prime\":           \"https:\u002F\u002Fwww.amazon.co.uk\u002Fmc\u002Fpipeline\u002FcancelEndBenefit\",\n  \"Disney+\":                \"https:\u002F\u002Fwww.disneyplus.com\u002Faccount\u002Fsubscription\",\n  \"Microsoft 365\":          \"https:\u002F\u002Faccount.microsoft.com\u002Fservices\",\n  \"GitHub\":                 \"https:\u002F\u002Fgithub.com\u002Fsettings\u002Fbilling\u002Fsubscriptions\",\n  \"Dropbox\":                \"https:\u002F\u002Fwww.dropbox.com\u002Faccount\u002Fplan\",\n  \"Zoom\":                   \"https:\u002F\u002Fzoom.us\u002Fbilling\",\n  \"LinkedIn Premium\":       \"https:\u002F\u002Fwww.linkedin.com\u002Fpremium\u002Fproducts\",\n  \"Google One\u002FWorkspace\":   \"https:\u002F\u002Fmyaccount.google.com\u002Fpayments-and-subscriptions\",\n  \"OpenAI\":                 \"https:\u002F\u002Fplatform.openai.com\u002Faccount\u002Fbilling\"\n}\n```\n\n---\n\n## Invocation\n\n```bash\n# Analyse a CSV file\npython3 subscription_killer.py --file transactions.csv\n\n# Override currency display\nSUBSCRIPTION_KILLER_CURRENCY=EUR python3 subscription_killer.py --file transactions.csv\n\n# Output raw JSON for piping to another skill\npython3 subscription_killer.py --file transactions.csv --json\n```\n\nThe script reads `SUBSCRIPTION_KILLER_CURRENCY` from the environment for display\nonly; all internal calculations use the amounts as-found in the CSV.\n\n---\n\n## Roadmap (future skills in this suite)\n\nThis skill is intentionally scoped to subscription detection. Planned companion\nskills that share the CSV parsing layer:\n\n| Skill | Description |\n|-------|-------------|\n| `savings-rate-analyser` | Track income vs spend, compute savings rate, compare to benchmarks |\n| `deposit-rate-scout`    | Pull live UK\u002FEU savings account rates via open banking APIs |\n| `emergency-fund-checker`| Assess months of runway based on average monthly spend |\n| `investment-nudger`     | Identify investable surplus after subscriptions + essentials |\n\nThe `bank-csv-parser` module within this skill will be extracted into a shared\nutility skill once the suite matures.\n\n---\n\n## Security & privacy\n\n- All processing is local. No transaction data leaves the machine.\n- The skill does not write any files outside the working directory.\n- No API keys required for the core subscription detection flow.\n- The `cancel_urls.json` is a static lookup table; no network calls are made\n  unless the user explicitly requests live rate data (future skill).\n","# Subscription Killer\n\nAnalyse a bank transactions CSV file and surface every recurring subscription,\nranked by cancellation \u002F downgrade priority with estimated monthly savings.\n\n## Skill overview\n\nThis skill takes a single CSV file of bank transactions and:\n\n1. Sniffs the CSV structure to detect column names automatically.\n2. Normalises merchant names using a curated alias dictionary plus fuzzy\n   matching (Levenshtein distance ≤ 2 tokens).\n3. Detects recurring charges by cadence (monthly \u002F quarterly \u002F annual) and\n   amount consistency (±2 % tolerance).\n4. Scores each subscription by confidence (0–100).\n5. Ranks subscriptions by a composite priority score: spend × inactivity risk ×\n   price-creep factor × duplication penalty.\n6. Outputs a structured report with clear cancellation and downgrade actions.\n\n---\n\n## Input format\n\nThe skill accepts any CSV export from a personal bank account. Common formats\nare auto-detected:\n\n| Bank \u002F Provider | Typical columns |\n|-----------------|-----------------|\n| Monzo | Date, Name, Amount, Category |\n| Starling | Date, Counter Party, Amount, Balance |\n| Revolut | Started Date, Description, Amount, Currency |\n| Standard OFX export | Date, Description, Debit Amount |\n| Generic | date, merchant \u002F description, amount (minimum required) |\n\nColumn detection priority:\n- **Date**: looks for columns named `date`, `transaction date`, `started date`,\n  `posted date` (case-insensitive). Falls back to first column that parses as\n  ISO 8601 or DD\u002FMM\u002FYYYY.\n- **Amount**: looks for `amount`, `debit amount`, `debit`. Ignores credit \u002F\n  top-up rows (positive amounts in debit-positive conventions).\n- **Merchant**: looks for `name`, `description`, `merchant`, `counter party`,\n  `payee`. Falls back to longest non-numeric column.\n\n---\n\n## Core logic\n\n### Step 1 — Parse CSV\n\n```python\nimport csv, re, difflib\nfrom datetime import datetime, timedelta\nfrom collections import defaultdict\n\ndef sniff_columns(header: list[str]) -> dict:\n    \"\"\"Return {'date': col, 'amount': col, 'merchant': col} best guesses.\"\"\"\n    header_lower = [h.lower().strip() for h in header]\n    DATE_HINTS    = ['date', 'transaction date', 'started date', 'posted date']\n    AMOUNT_HINTS  = ['amount', 'debit amount', 'debit']\n    MERCHANT_HINTS = ['name', 'description', 'merchant', 'counter party',\n                      'payee', 'narrative']\n    def first_match(hints):\n        for hint in hints:\n            if hint in header_lower:\n                return header[header_lower.index(hint)]\n        return None\n    return {\n        'date':     first_match(DATE_HINTS)     or header[0],\n        'amount':   first_match(AMOUNT_HINTS)   or header[1],\n        'merchant': first_match(MERCHANT_HINTS) or header[2],\n    }\n\ndef parse_amount(raw: str) -> float | None:\n    \"\"\"Return a positive debit amount, or None for credits \u002F errors.\"\"\"\n    cleaned = re.sub(r'[£$€,\\s]', '', raw.strip())\n    if not cleaned:\n        return None\n    try:\n        val = float(cleaned)\n        # Debit-positive CSV (most UK banks): only keep debits (negatives after\n        # sign flip are credits — skip them)\n        # Some CSVs use negative for debits; normalise to positive debit.\n        return abs(val) if val != 0 else None\n    except ValueError:\n        return None\n```\n\n### Step 2 — Normalise merchant names\n\nUse a two-pass approach:\n\n**Pass 1 — Known alias dictionary** (curated, highest-confidence)\n\n```python\nMERCHANT_ALIASES = {\n    # Streaming\n    r'netflix':                 'Netflix',\n    r'spotify':                 'Spotify',\n    r'apple\\.com\u002Fbill':         'Apple Subscriptions',\n    r'itunes':                  'Apple Subscriptions',\n    r'amazon prime':            'Amazon Prime',\n    r'amzn\\s?prime':            'Amazon Prime',\n    r'disney\\+|disneyplus':     'Disney+',\n    r'youtube premium':         'YouTube Premium',\n    r'hbo|max\\.com':            'Max (HBO)',\n    r'paramount':               'Paramount+',\n    r'dazn':                    'DAZN',\n    # Music\n    r'tidal':                   'Tidal',\n    r'deezer':                  'Deezer',\n    r'soundcloud':              'SoundCloud',\n    # Productivity \u002F SaaS\n    r'github':                  'GitHub',\n    r'notion':                  'Notion',\n    r'slack':                   'Slack',\n    r'dropbox':                 'Dropbox',\n    r'google\\s?(one|storage|workspace)': 'Google One\u002FWorkspace',\n    r'microsoft 365|office 365|msft': 'Microsoft 365',\n    r'adobe':                   'Adobe Creative Cloud',\n    r'figma':                   'Figma',\n    r'zoom':                    'Zoom',\n    r'1password|lastpass|bitwarden': 'Password Manager',\n    r'nordvpn|expressvpn|surfshark':  'VPN Service',\n    # News \u002F Reading\n    r'medium':                  'Medium',\n    r'substack':                'Substack',\n    r'kindle unlimited':        'Kindle Unlimited',\n    r'audible':                 'Audible',\n    r'the times|thetimes':      'The Times',\n    r'financial times|ft\\.com': 'Financial Times',\n    r'economist':               'The Economist',\n    # Fitness\n    r'peloton':                 'Peloton',\n    r'strava':                  'Strava',\n    r'myfitnesspal':            'MyFitnessPal',\n    r'calm|headspace':          'Meditation App',\n    # Utilities \u002F services\n    r'amazon web services|aws': 'AWS',\n    r'google cloud|gcp':        'Google Cloud',\n    r'digitalocean':            'DigitalOcean',\n    r'cloudflare':              'Cloudflare',\n    r'railway\\.app':            'Railway',\n    r'openai':                  'OpenAI',\n    r'anthropic':               'Anthropic',\n}\n\ndef normalize_merchant(raw: str) -> str:\n    \"\"\"Return a clean, canonical merchant name.\"\"\"\n    s = raw.lower().strip()\n    for pattern, canonical in MERCHANT_ALIASES.items():\n        if re.search(pattern, s, re.IGNORECASE):\n            return canonical\n    # Pass 2 — strip noise tokens (card numbers, references, country codes)\n    cleaned = re.sub(\n        r'\\b(www|com|co\\.uk|ltd|limited|uk|us|gb|inc|gmbh)\\b'\n        r'|\\d{4,}'      # card \u002F ref numbers\n        r'|\\*+\\S*'      # asterisk-prefixed tokens\n        r'|[^a-z0-9 ]', # special chars\n        '', s\n    ).strip().title()\n    return cleaned or raw.title()\n```\n\n**Pass 2 — Fuzzy deduplication** across the transaction list\n\nAfter normalisation, cluster remaining merchant names by token-set similarity\n(`difflib.SequenceMatcher` ratio > 0.82). The most-frequent variant in each\ncluster becomes the canonical name.\n\n### Step 3 — Detect recurring charges\n\n```python\ndef detect_cadence(dates: list[datetime]) -> str | None:\n    \"\"\"\n    Return 'monthly', 'quarterly', 'annual', or None.\n    Requires ≥ 2 transactions for monthly\u002Fquarterly, ≥ 1 for annual hinting.\n    \"\"\"\n    if len(dates) \u003C 2:\n        # Single occurrence — flag as potential annual if amount > 30\n        return 'annual_candidate'\n    gaps = sorted([(dates[i+1] - dates[i]).days for i in range(len(dates)-1)])\n    median_gap = gaps[len(gaps)\u002F\u002F2]\n    if 25 \u003C= median_gap \u003C= 35:   return 'monthly'\n    if 85 \u003C= median_gap \u003C= 100:  return 'quarterly'\n    if 340 \u003C= median_gap \u003C= 380: return 'annual'\n    return None\n\ndef amounts_consistent(amounts: list[float], tolerance: float = 0.02) -> bool:\n    \"\"\"True if all amounts are within ±2 % of the median.\"\"\"\n    if not amounts:\n        return False\n    median = sorted(amounts)[len(amounts)\u002F\u002F2]\n    return all(abs(a - median) \u002F median \u003C= tolerance for a in amounts)\n```\n\n### Step 4 — Confidence scoring (0–100)\n\n| Factor | Max points | Logic |\n|--------|-----------|-------|\n| Cadence regularity | 40 | All gaps within ±3 days of median |\n| Amount consistency | 30 | All amounts within ±2 % |\n| Known merchant match | 20 | Exact alias dictionary hit |\n| Transaction count | 10 | log₂(count) × 3, capped at 10 |\n\nConfidence ≥ 70 → **Confirmed subscription**\nConfidence 40–69 → **Probable subscription**\nConfidence \u003C 40 → **Possible \u002F flag for review**\n\n### Step 5 — Priority ranking\n\nEach confirmed\u002Fprobable subscription gets a composite priority score:\n\n```\npriority = monthly_cost\n         × price_creep_multiplier   # max(amounts) \u002F min(amounts); 1.0 = stable\n         × duplication_penalty       # 1.5 if same category has 2+ subscriptions\n         × inactivity_risk           # 1.3 if last charge > 45 days ago\n```\n\nHigher priority score = stronger cancellation candidate.\n\n**Special flags:**\n- **Trial trap**: single charge in the last 7 days from a previously-unseen\n  merchant, followed by a recurring pattern. Flag as \"check for auto-renew\".\n- **Price creep**: amount has increased > 5 % over the observation window.\n  Flag with the exact amount delta.\n- **Annual renewal due**: annual-cadence subscription whose last charge was\n  300–365 days ago. Flag with estimated renewal date.\n- **Unknown merchant**: low-confidence recurring charge with no alias match.\n  Flag for manual review.\n\n---\n\n## Output schema\n\nThe skill returns a structured report. Print to stdout as JSON for downstream\nagent consumption, or render the human-readable summary below.\n\n```json\n{\n  \"summary\": {\n    \"observation_window_days\": 90,\n    \"total_subscriptions\": 12,\n    \"confirmed\": 9,\n    \"probable\": 2,\n    \"flagged_for_review\": 1,\n    \"monthly_spend_gbp\": 187.43,\n    \"annual_spend_gbp\": 2249.16,\n    \"potential_monthly_saving_gbp\": 54.20\n  },\n  \"subscriptions\": [\n    {\n      \"merchant\": \"Adobe Creative Cloud\",\n      \"cadence\": \"monthly\",\n      \"median_amount\": 54.99,\n      \"last_charge\": \"2025-04-01\",\n      \"confidence\": 95,\n      \"priority_score\": 89.2,\n      \"flags\": [\"price_creep\"],\n      \"price_creep_delta\": 5.00,\n      \"cancel_url\": \"https:\u002F\u002Faccount.adobe.com\u002Fplans\",\n      \"actions\": [\"cancel\", \"downgrade\"]\n    }\n  ],\n  \"top_cancellation_targets\": [\"Adobe Creative Cloud\", \"...\"],\n  \"unknown_merchants\": [\"TRSF 48291 REF77\"],\n  \"annual_renewals_due\": []\n}\n```\n\n### Human-readable summary (terminal \u002F Telegram)\n\n```\n╔══════════════════════════════════════════╗\n║        SUBSCRIPTION KILLER  v1.0         ║\n╠══════════════════════════════════════════╣\n║  12 subscriptions  •  £187\u002Fmo  •  £2,249\u002Fyr\n╚══════════════════════════════════════════╝\n\nTOP CANCELLATION TARGETS\n1. Adobe Creative Cloud   £54.99\u002Fmo  ↑ price crept +£5.00\n2. Calm                   £39.99\u002Fyr  last used 3 months ago\n3. LinkedIn Premium       £29.99\u002Fmo  duplicate: 2 job-search tools\n\nPOTENTIAL SAVING:  £54\u002Fmo  •  £648\u002Fyr\n\nUNKNOWN MERCHANTS (review manually)\n• TRSF 48291 REF77  — £12.99\u002Fmo  (confidence 38%)\n```\n\n---\n\n## Cancel guidance (top 50 services)\n\nThe skill bundles a `cancel_urls.json` file mapping canonical merchant names\nto their self-serve cancellation pages. This removes friction for the user.\n\n```json\n{\n  \"Netflix\":                \"https:\u002F\u002Fwww.netflix.com\u002Fcancelplan\",\n  \"Spotify\":                \"https:\u002F\u002Fwww.spotify.com\u002Faccount\u002Fsubscription\u002Fcancel\",\n  \"Adobe Creative Cloud\":   \"https:\u002F\u002Faccount.adobe.com\u002Fplans\",\n  \"Amazon Prime\":           \"https:\u002F\u002Fwww.amazon.co.uk\u002Fmc\u002Fpipeline\u002FcancelEndBenefit\",\n  \"Disney+\":                \"https:\u002F\u002Fwww.disneyplus.com\u002Faccount\u002Fsubscription\",\n  \"Microsoft 365\":          \"https:\u002F\u002Faccount.microsoft.com\u002Fservices\",\n  \"GitHub\":                 \"https:\u002F\u002Fgithub.com\u002Fsettings\u002Fbilling\u002Fsubscriptions\",\n  \"Dropbox\":                \"https:\u002F\u002Fwww.dropbox.com\u002Faccount\u002Fplan\",\n  \"Zoom\":                   \"https:\u002F\u002Fzoom.us\u002Fbilling\",\n  \"LinkedIn Premium\":       \"https:\u002F\u002Fwww.linkedin.com\u002Fpremium\u002Fproducts\",\n  \"Google One\u002FWorkspace\":   \"https:\u002F\u002Fmyaccount.google.com\u002Fpayments-and-subscriptions\",\n  \"OpenAI\":                 \"https:\u002F\u002Fplatform.openai.com\u002Faccount\u002Fbilling\"\n}\n```\n\n---\n\n## Invocation\n\n```bash\n# Analyse a CSV file\npython3 subscription_killer.py --file transactions.csv\n\n# Override currency display\nSUBSCRIPTION_KILLER_CURRENCY=EUR python3 subscription_killer.py --file transactions.csv\n\n# Output raw JSON for piping to another skill\npython3 subscription_killer.py --file transactions.csv --json\n```\n\nThe script reads `SUBSCRIPTION_KILLER_CURRENCY` from the environment for display\nonly; all internal calculations use the amounts as-found in the CSV.\n\n---\n\n## Roadmap (future skills in this suite)\n\nThis skill is intentionally scoped to subscription detection. Planned companion\nskills that share the CSV parsing layer:\n\n| Skill | Description |\n|-------|-------------|\n| `savings-rate-analyser` | Track income vs spend, compute savings rate, compare to benchmarks |\n| `deposit-rate-scout`    | Pull live UK\u002FEU savings account rates via open banking APIs |\n| `emergency-fund-checker`| Assess months of runway based on average monthly spend |\n| `investment-nudger`     | Identify investable surplus after subscriptions + essentials |\n\nThe `bank-csv-parser` module within this skill will be extracted into a shared\nutility skill once the suite matures.\n\n---\n\n## Security & privacy\n\n- All processing is local. No transaction data leaves the machine.\n- The skill does not write any files outside the working directory.\n- No API keys required for the core subscription detection flow.\n- The `cancel_urls.json` is a static lookup table; no network calls are made\n  unless the user explicitly requests live rate data (future skill).\n",{"name":5,"version":24,"metadata":25,"description":35},"1.1.0",{"openclaw":26},{"envVars":27,"requires":31,"primaryEnv":29},[28],{"name":29,"required":20,"description":30},"SUBSCRIPTION_KILLER_CURRENCY","ISO 4217 currency code for display (default: GBP). E.g. USD, EUR, GBP.\n",{"env":32,"bins":33},[],[34],"python3","Analyse a bank transactions CSV to detect recurring subscriptions, score cancellation priority, and surface actionable savings recommendations. Designed as the first module in a broader personal-finance agent; bank data parsing is deliberately decoupled so future savings \u002F deposit \u002F investment skills can reuse it.\n","master","4aa004da461d820f4ea5499e58eb8d196cf99016","community\u002Fclawhub\u002Fs\u002Fsubscription-killer",0,"https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge","https:\u002F\u002Fskill-explorer-rho.vercel.app","https:\u002F\u002Fgithub.com\u002FLord1Egypt","Unbreakable. Unstoppable. Unmatched⚡.\r\n Speaker of machine tongues 🤖  | World-builder of decentralized dreams 🌐 | Defender of the digital realm🛡️.","2026-06-10 08:21:01",{"source":8,"skillName":5,"commands":46,"prompt":56,"zipAvailable":57},[47,50,53],{"tool":48,"command":49},"npx","npx skills add Lord1Egypt\u002Fawesome-skill-forge --skill subscription-killer",{"tool":51,"command":52},"bunx","bunx skills add Lord1Egypt\u002Fawesome-skill-forge --skill subscription-killer",{"tool":54,"command":55},"pnpm","pnpm dlx skills add Lord1Egypt\u002Fawesome-skill-forge --skill subscription-killer","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fcommunity\u002Fclawhub\u002Fs\u002Fsubscription-killer\u002FSKILL.md",true,[],[60,67,74,81,88,95],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M785WKJB99G3STR1NF9NRTJW","youtube-content","YouTube transcripts to summaries, threads, blogs.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fyoutube-content\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M785WEFHB0G2S9YA6EPR3H0E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fxurl\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M785W98EDN4EFJXJPMQEHCRC","touchdesigner-mcp","Control a running TouchDesigner instance via twozero MCP — create operators, set parameters, wire connections, execute Python, build real-time visuals. 36 native tools.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftouchdesigner-mcp\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M785W5S9FP3QBH2V191E025P","test-driven-development","TDD: enforce RED-GREEN-REFACTOR, tests before code.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Ftest-driven-development\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M785W2AS08NSSH9YATCV6FJB","teams-meeting-pipeline","Operate the Teams meeting summary pipeline via Hermes CLI — summarize meetings, inspect pipeline status, replay jobs, manage Microsoft Graph subscriptions.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fteams-meeting-pipeline\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":19},"01M785VXKW4T34BTP1SCHW1R7F","systematic-debugging","4-phase root cause debugging: understand bugs before fixing.","https:\u002F\u002Fgithub.com\u002FLord1Egypt\u002Fawesome-skill-forge\u002Fblob\u002Fmaster\u002Fskills\u002Fsystematic-debugging\u002FSKILL.md",[],[]]