simon gant

    0followers
    13 skills
    0 uses
    Joined Nov 2026

    Skills by simon gant (13)

    Provable AI Output Verifiers

    by simon gant

    $29

    Six Agent Trust skills in one pack: hidden-Unicode audit, quote/citation verifier, JSON Schema gate, agent spend audit, canonical JSON diff and near-duplicate check. Each returns a checkable verdict and SHA-256 receipt, not an opinion. Offline, zero dependencies. $42 if bought separately.

    1
    0

    Test Theatre Gate: Mutation & Fuzz Check

    by simon gant

    $15

    Green tests can lie. Flips one operator at a time in your source file, reruns your own test command, and lists every line where the tests still pass, i.e. code your tests don't really check. Also fuzzes an exported function with hostile inputs. Always restores your file.

    1
    0

    Agent Tool Trace Audit

    by simon gant

    $15

    Check what an AI agent actually did against a written policy: allowed and forbidden tools, required calls, output JSON shape, numeric limits, leaked strings and a step ceiling. Names every violation with its step. No LLM judge. Hash-chained audit ledger and a PASSES/FAILS suite receipt.

    1
    0

    Skill & Prompt Drift Pin

    by simon gant

    $9

    Pin an agent skill or prompt folder once, then catch any change: edited frontmatter, a new tool grant, new network, exec, file-write or env access in its code, a changed or missing file, or hidden Unicode. Returns INTACT, DRIFTED or BREACHED with a SHA-256 receipt anyone can re-run.

    1
    0

    Redaction Proof

    by simon gant

    $12

    Redact emails, phone numbers, NI numbers, card numbers, postcodes or your own patterns from a text file, and give the recipient a proof they can check without the original: a SHA-256 Merkle root over the lines plus a seal over what was removed. Scans the output for anything left behind.

    1
    0

    Claim Receipt Gate

    by simon gant

    $12

    Stop a release note or launch post from claiming more than you measured. Pulls every score, percentage, money amount, multiplier, URL and live/CI/verified claim from a draft and checks each against an evidence file. Returns SHIP or REFUSE with the exact words and line, plus a SHA-256 seal.

    1
    0

    Copy Gate: Six Layer Copy Review

    by simon gant

    $12

    Score a draft before it ships. Six weighted lenses (archetype voice, motive, sales structure, SEO, AI-search citability, platform fit) are tuned to the channel: sales page, LinkedIn, email, README and more. You get a 0–100 gate score plus findings ranked by impact. Rule-based and offline.

    1
    0

    Agent Spend Ledger Audit

    by simon gant

    $7

    Recompute an agent's spend ledger line by line from the budget cap. Flags overspend, skipped sequence numbers and claimed balances that don't add up, using integer-cent maths with no float drift. Returns PASS/FAIL, claimed vs recomputed balances and a SHA-256 fingerprint.

    1
    0

    Near Duplicate Text Detector

    by simon gant

    $5

    Is this draft a lightly reworded copy? Compares two texts with word shingles and a deterministic MinHash signature and returns IDENTICAL, NEAR-DUPLICATE, OVERLAPPING or DISTINCT, with a similarity score and a SHA-256 receipt. Offline, no AI judge.

    1
    0

    Canonical JSON Hash & Diff

    by simon gant

    $5

    Prove two JSON documents are identical whatever their key order, spacing or number formatting, or get a path-by-path list of what was added, removed, changed or changed type. Canonical SHA-256 hashes you can quote as a receipt. Deterministic and offline.

    1
    0

    JSON Schema Conformance Gate for LLM Outputs

    by simon gant

    $7

    Validate an agent's JSON output against a JSON Schema before your code trusts it. Lists every violation by path, gives each output a canonical SHA-256 fingerprint, and groups runs that are the same content despite key order or whitespace. Deterministic and offline.

    1
    0

    Quote & Citation Verifier

    by simon gant

    $9

    Checks every quote in an AI draft against the source text and returns VERIFIED, NEAR_MATCH or NOT_FOUND per quote, with the surrounding excerpt and a SHA-256 receipt. Catches invented and misworded quotes before you publish. Deterministic and offline.

    1
    0

    Hidden Unicode & Prompt Smuggling Audit

    by simon gant

    $9

    Before your agent acts on pasted text, check it for invisible Unicode tag characters, bidi overrides, look-alike letters and zero-width characters. You get a flagged/clean verdict, the decoded smuggled text and a SHA-256 receipt. Offline and deterministic.

    1
    0