Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Independent Cross-Check: verify 'done' a second way

    1

    A short plain-text skill that makes your coding agent check any 'done / OK / verified' by a different route than the one that produced it, and leave one line of proof.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Independent Cross-Check: verify 'done' a second way

    Independent Cross-Check: verify 'done' a second way

    Example session with this skill installed

    I've updated the regex in parser.py to handle ISO dates. Before you tell me it's done, use the independent-cross-check skill.

    • Read your context and instructions
    • Compiled the independent cross-check

    Cross-check: ISO date regex update (parser.py)

    Illustrative example written by us to show the format; your output will depend on your project.

    Claim (could be false): "parse_date() returns the right date for every ISO 8601 string in tests/fixtures/dates.txt on Python 3.9 and 3.12."

    Different route: ran the 40 fixture strings through the new function in a fresh interpreter and compared each result with datetime.fromisoformat, instead of re-reading my own tests.

    Result: 38 of 40 match. Two differ: "2026-10-30T10:00Z" (trailing Z) and "2026-W44-5" (week date).

    Pre-mortem (3 lines):

    • Python 3.9 fromisoformat does not accept a trailing Z, and production runs 3.9.
    • Naive and timezone-aware values may compare wrongly around a clock change.
    • Empty-string input is not covered by any fixture.

    Verdict: NOT done. Fix the Z handling; decide whether week dates are in scope.

    Trail (one line): "Checked parse_date on 40 fixtures in a fresh interpreter against fromisoformat: 2 mismatches, fixing Z."

    Next steps: add the two failing strings and an empty string to the fixtures, re-run on both Python versions, then report again.

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

    About this skill

    What it does When Claude Code (or any agent that reads SKILL.md) says "done", this skill makes it check that claim a second way before you believe it.

    It asks the agent to:

    1. Write the claim as one sentence that could turn out false.
    2. Check it by a route other than the one that produced it (another interpreter, another data source, a clean re-run, the real state), not by re-reading its own report.
    3. Write a 3-line pre-mortem: in what different context would this fail?
    4. For analysis figures, look for survivorship, look-ahead and ignored costs, and compare against chance.
    5. Pick the cheapest check that could refute the claim before launching more agents.
    6. Leave one line saying what was checked and how.
    7. Stay read-only: local tests and simulations only, never anything irreversible.

    What you get A folder with SKILL.md (the rules, about 1.5 KB) and WHY.md (the real problem it came from). Plain text: open it and read it before you install it.

    Who it is for People who use coding agents for real work and keep finding "validated" tasks that nobody re-ran.

    Honest limits It is a set of instructions, not software. It changes how the agent works; it does not guarantee correct results. Written from our own routine and tried in Claude Code. We have not run it inside Cursor, Codex or Gemini CLI. Independent; not affiliated with Anthropic.

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

    Read-only and local simulation scope only. Cannot execute irreversible actions or production writes for verification purposes.

    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

    Money-back guarantee

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

    Trust & safety

    Security scan

    Security scan passed 18 days ago

    Permissions

    Terminal / Shell
    Read Files
    Listed18 days ago

    What's inside

    Frequently Asked Questions

    A plain 'check your work' often just re-reads the same answer. This skill asks for a check by a different route (another interpreter, another data source, a clean re-run), a 3-line pre-mortem, and one line of proof.

    It is a SKILL.md file, so any agent that reads SKILL.md can load it. We have tried it in Claude Code. We have not run it inside Cursor, Codex or Gemini CLI.

    The skill only gives instructions. It tells the agent to stay read-only: local tests, greps and dry runs, never anything irreversible. What the agent actually runs is still governed by your own permission settings.

    A zip with two plain-text files: SKILL.md (the rules, about 1.5 KB) and WHY.md (the problem it came from). You can read both before installing.

    No. It is a habit written down for the agent. It makes unchecked 'done' claims harder to slip through, but you still decide what counts as verified.

    New in Testing & Debugging