Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Fable Receipts

    1

    The working rules Claude Fable 5.2 wrote for Opus 5.5. 17 rules, each tied to the real mistake that created it.

    Free

    Security scanned
    fable-receipts

    fable-receipts

    Example session with this skill installed

    Create the GitHub repo for my new tool and set its description and topics.

    • Read your context and instructions
    • Compiled the fable-receipts

    Repo created: MBotirjonDev/fable-receipts. CI is green.

    Verified, not assumed:

    • I read the description back through the API after saving. The first version had lost characters while typing ("Opus 5.." and "crated"). I set it again and read it back: now correct. (R11: reload and look before saying "saved")
    • Topics: 7 saved, confirmed through the API.
    • Social preview: reloaded the settings page, image is shown.

    Not done: release and marketplace listing. Want me to publish the v0.1.0 release now?

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

    About this skill

    I used Claude Fable 5.2 for a few days and shipped three open-source projects with it (a Dart package on pub.dev and two agent skills). Before switching back to Opus 5.5, I asked Fable 5.2 to write down how it works so Opus could follow the same discipline.

    What makes this skill different: it does not describe an ideal process. Every rule points to a receipt, a real incident from that session. For example: a linter found 11 invisible Unicode characters in its own source; a pub.dev setting did not save because a confirm dialog was never accepted; 36 screenshots ended up costing about 2.7M tokens because they stayed in context.

    What the agent does with it:

    • Three gates before code: check what exists, search the name and competition, write the one sentence that makes the project different.
    • Verify like a skeptic: tests, own examples, 3+ real outside projects with every finding read, self-lint, version constraints, every claim checked against evidence.
    • Truth rules: only real numbers, "saved" only after a reload, say what was not tested.
    • Asks before publishing, posting, paying or deleting.
    • Usage-limit habits: text over screenshots, batched actions, save first.
    • Short reports: what came out, what was verified, what was not.

    Honest limits: a skill changes habits, not capability. It will not make Opus think like Fable. No benchmark yet. Free and MIT on GitHub: github.com/MBotirjonDev/fable-receipts

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.

    Cannot handle secrets, passwords, or account creation. Requires human approval for all irreversible public actions. Does not perform actual web searches if tools are restricted.

    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 buying a skill.

    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

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 6 days ago

    • Free with an Agensi account

    Requires

    Nothing to install.

    Permissions

    Read Files
    Terminal / Shell
    Write Files
    Listed6 days ago
    Updated2 days ago

    What's inside

    Frequently Asked Questions

    It gives the agent concrete checks before it may say "done", "saved" or "published": run the tests, run the tool on real outside projects and read every finding, reload the page after a save, and re-read every claim against the evidence. Each check comes with the real incident that made it necessary.

    Any agent that reads SKILL.md files: Claude Code, Codex CLI, Cursor, OpenCode and others. It is plain Markdown with no scripts, so there is nothing to install.

    No. A skill changes habits, not capability. It stops the mistakes that make work look finished when it is not: invented numbers, unverified saves, untested claims. I have not benchmarked it yet.

    It tells the agent to read pages as text instead of screenshots, batch actions, keep command output short, and save or commit before polishing, so a limit hit mid-task does not lose work.

    SKILL.md with the rules and references/receipts.md with 17 real incidents behind them. The same skill is free under MIT on GitHub (MBotirjonDev/fable-receipts); buying here supports the work and gives you the packaged version.

    New in Agents & Orchestration

    Free