Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Post Receipts

    1

    Never trust a "posted" that never went out: your agent reports success only after reading the live URL back and matching the text.

    $12

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Post Receipts

    Post Receipts

    Example session with this skill installed

    Post this reply on X under the patch-notes thread: "patch notes for the new season are up on battle.net and the ranked reset finally landed, so every player who grinded to the top last month starts from zero again this week, and queue times tonight are going to be brutal for anyone climbing before qualifiers this weekend"

    • Read your context and instructions
    • Compiled the post receipts

    Pre-flight

    $ receipts.py pending
    nothing pending
    $ receipts.py check --platform x "patch notes for the new season ... qualifiers this weekend"
    x: 283 / 280  (1 link(s) counted as 23 each)  TOO LONG by 3
    

    The text is 270 characters, but X turns "battle.net" into a link and counts it as 23, so it is 283. Posting it would fail silently. I trimmed "this weekend":

    x: 270 / 280  (1 link(s) counted as 23 each)  OK
    $ receipts.py intend --platform x "..."
    7c2d90a1e4
    

    Posted, then checked

    The post page needs JavaScript, so I read the reply back from the profile's replies tab and matched it:

    $ receipts.py confirm 7c2d90a1e4 --url https://x.com/<account>/status/<id> --evidence "patch notes for the new season are up on battle.net and the ranked reset..."
    MATCH in read-back evidence (links ignored, platforms rewrite them): "patch notes for the new season are up on and the ranked reset finally landed, so"
    POSTED 7c2d90a1e4 -> https://x.com/<account>/status/<id>
    

    Result: posted, with receipt. 1 intent, 1 posted, 0 pending, 0 retries.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Outcome: your posting agent can only say "posted" with a receipt, the live URL read back and the text matched, so your log never lists a post that isn't there.

    Why this over trusting the agent: agents report success when the click happened, not when the post went live. This checks the live page.

    Guaranteed: if it doesn't do what this page says on your setup, message me here or email support@tensorust.com and I fix it.

    What it prevents

    Posting agents report success they never had. The script returns a URL, the button looked enabled, the run log says "posted", and nothing is on the platform. Or the opposite: the script crashes after the post went through, the agent runs it again, and the post goes out twice. Or a post gets removed and the agent reposts it somewhere else right away, which is how accounts get filtered.

    The incident behind it (true)

    From one team's logs while running agents that post to social platforms:

    • On X, the posting script typed posts over 280 characters. Nothing was posted, and the script returned the URL of the previous post as success. It happened more than once, and a reply script printed "replied" on a 293-character reply that never sent.
    • A 269-character reply contained "battle.net". X auto-links bare domains and counts each link as 23 characters, so it was over 280 and never sent, while len() said 269.
    • A script posting to an agent forum crashed while printing the response, after the request had gone through. The status said failed; the platform said posted.
    • On Reddit, removed posts were reposted to another community right after, three times or more. That pattern preceded every post being filter-removed and a site-wide ban (later lifted).

    What you get

    1. The rule: a post counts only with a receipt (live URL read back, text matched), with a table of what the agent may say in every other case ("attempted, unverified", "unknown, pending").
    2. Pre-flight length checks in receipts.py check: X weighted length per twitter-text v3 (URLs and bare domains 23, CJK and emoji 2, emoji sequences counted once), plus LinkedIn post/comment, Bluesky (graphemes), Threads, Reddit title and Mastodon, each overridable with --limit where the limit is uncertain.
    3. Write-ahead journal: intend logs the post before it is sent, confirm attaches the receipt after. On restart, pending lists every unresolved post; the agent checks the platform before anything else. intend refuses a text that is still pending or already posted.
    4. No retry chains: retry only after the first attempt is confirmed absent, max 1 retry, never move a removed post to another venue the same day.
    5. verify URL "text": fetches a public page and looks for the text (links stripped, HTML and whitespace normalized). Offline or blocked returns "could not check", never a false receipt. For JavaScript-only pages, confirm --evidence matches text the agent read back itself.
    6. File-locked JSONL safe for several agents and cron jobs at once, crash-tolerant, standard library only, with a test suite (25 tests: length math incl. URL/emoji/CJK cases, crash reconciliation, retry cap, concurrent writers, offline).

    Honest limit: the X URL detector uses a common-TLD list, not every TLD, and platforms change their rules. When the platform disagrees, the platform wins.

    Verification fetches public URLs only; pages behind a login or heavy JavaScript need the agent to read the post back itself and pass the text to confirm. X length rules follow twitter-text v3 as implemented here, not the official library; bare-domain detection uses a list of common endings. Limits for LinkedIn comments, Threads and Mastodon vary and are configurable.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Permissions

    Terminal / Shell
    Write Files
    Network Access
    Environment Variables

    Runs receipts.py locally (writes one JSONL journal file). Network only to fetch the public post URL you pass to verify/confirm.

    Listed1 month ago
    Updated26 days ago

    What's inside

    Frequently Asked Questions

    This skill provides a failsafe layer that prevents double-posting after crashes and uses hard-coded logic to catch character-count errors (like link-weighting on X) that LLMs often miscalculate.

    The skill is a standalone Python 3 utility that works with any agent framework or custom script, as long as your agent can execute local Python functions and handle JSONL files.

    You receive the core `receipts.py` script, documentation for the pre-flight and journaling logic, and a guide on integrating the verification workflow into your existing posting scripts.

    It supports X (Twitter), LinkedIn, Bluesky, Threads, Reddit, and Mastodon, with platform-specific rules for character scaling and URL weighting built directly into the checker.

    Verification is performed via the `confirm` function, which fetches the live URL provided by the platform API and matches the page content against your agent's intended text.

    Every post is logged as an intent before the agent acts. On restart, `receipts.py pending` lists anything not yet confirmed; the agent checks the platform for each one and marks it posted or absent. Only an attempt confirmed absent may be retried, and only once.

    New in Social Media