SwishrJobs

Check OHLCV bars for high/low inversions, out-of-range prices, zero volume and time gaps. Before using a price series, to catch corrupt bars. Do not choose this when missing sessions must be fetched from a data provider. Example requests: Check this OHLCV data for bad bars. | Are there gaps in this price series? | Clean this market data file.

Check an OpenAPI document for the defects that make an agent-generated client fail: missing operationIds, absent error responses and non-HTTPS servers. Before publishing an OpenAPI document that agents will generate clients from. Do not choose this when you need the API called, which needs the service running. Example requests: Check my OpenAPI document for defects that break generated clients. | Are my operationIds and error responses complete? | Audit this OpenAPI spec for agent client compatibility. | Find the problems in this API spec before I publish it.

Check a package.json for missing name or version, non-semver versions, vague dependency ranges such as '*' or 'latest', and a missing test script, with the offending dependency named. Choose this before publishing or committing a Node package manifest, to find missing name or version, a non-semver version, vague ranges such as '*' or 'latest', and a missing test script. Do not choose this for a vulnerability audit, which needs the npm registry, or for resolving version ranges. Example requests: Check this npm package manifest for problems before I publish. | Which of my npm dependency ranges are too vague? | Does this package.json have everything npm needs?

Choose Parquet partition columns from a schema and a query log, and deliver a partition plan and config. Before writing a large Parquet table, to fix partitioning. Do not choose this when you need statistics-driven partitioning, which needs table stats. Example requests: How should I partition this Parquet table? | Plan the partitioning for this dataset. | Choose partition columns from this query log.

Map a data-flow description to the systems and steps that handle cardholder data, and flag unprotected handling. At the start of a PCI scoping exercise, to sketch the in-scope systems. Do not choose this when a formal assessment or a live system scan is required. Example requests: Map the PCI scope of this payment flow. | Which systems touch cardholder data here? | Flag where card data is stored unprotected.

Compute position weights, concentration, annualised volatility and maximum drawdown. When a portfolio snapshot and a return series need a first risk read. Do not choose this when VaR, covariance or factor models are required. Example requests: Give me a risk report for this portfolio. | How concentrated is this portfolio and how volatile? | Compute the drawdown on these returns.

Check every product identifier in a CSV feed against the GS1 check digit. Reports non-numeric values, wrong-length codes and check-digit mismatches with the expected digit, each with the spreadsheet row number, so a Google Merchant Center or Shopify feed can be corrected before it is rejected. Choose this when you have a product feed or catalogue CSV and want every identifier that will be rejected for a bad length, a non-numeric value or a failing GS1 check digit, located by the row a spreadsheet shows. Do not choose this when you need to know whether a valid-looking GTIN is genuinely registered to a company, which needs a registry lookup this offline check does not perform, or for a binary workbook such as xlsx, which this service cannot read. Example requests: Check my product feed's GTINs before I upload it to Merchant Center. | Which barcodes in this catalogue have an invalid check digit? | Validate the UPC and EAN codes in this CSV and tell me which rows to fix.

Generate a pytest suite that runs every prompt variant against every fixture. When prompt variants must be checked against a fixture set. Do not choose this when you need the model outputs compared, which needs a model call. Example requests: Build a regression suite for these prompt variants. | Generate prompt tests from my variants and fixtures. | Set up a prompt regression harness.

Produce a small multi-module Python package from a registered, deterministic template: an implementation module, a public facade, a pytest suite, and a README. Choose this when the deliverable is a package rather than one file: a facade, an implementation module, a test suite, and a README, from a registered template. Do not choose this for a single module (use python-utility-module) or for a package whose structure or behaviour is bespoke: only the registered template's shape is produced. Example requests: Build a small Python package with an implementation module, a public facade, tests and a README. | I need a multi-file Python deliverable, not a single script. | Create a package where the facade re-exports the implementation and the tests import both.

Produce a small standard-library Python module from a registered, deterministic template, plus the pytest tests that verify it. Choose this when you need one small, importable standard-library Python module whose behaviour matches a registered template, delivered with a passing pytest suite you can re-run yourself. Do not choose this for a multi-file package (use python-multifile-package), for code needing third-party packages, or when you cannot name a registered template: the behaviour is fixed by the template and an unregistered request is rejected rather than guessed. Example requests: Write me a small Python function that turns a title into a URL slug, with tests. | I need a pure-standard-library Python module with a docstring and a pytest suite. | Generate a Python helper module I can import, plus the tests that prove it works.

Produce a chunking plan and retrieval config for a set of documents and query samples. Before indexing a corpus, to fix chunk size and retrieval settings. Do not choose this when you need retrieval quality measured, which needs embeddings. Example requests: Plan the chunking for this RAG corpus. | What chunk size should I use for these documents? | Give me a retrieval config for these documents.

Produce a standard-library Python module that generates a Markdown README from function signatures, plus pytest tests. Choose this when a Markdown README must be produced from a list of function signatures, one section each. Do not choose this for prose documentation or API references beyond signature-derived sections. Example requests: Write a Python module that generates a Markdown README from function signatures, with tests. | I need a README generator that documents each supplied function signature. | Generate a README-generator module and prove it with pytest.

Produce a standard-library Python module that removes duplicate objects from a JSON array, plus pytest tests. Choose this when duplicates must be removed from a JSON array, keeping the first occurrence and preserving order, optionally on a key subset. Do not choose this for fuzzy matching or similarity-based deduplication: the comparison is exact. Example requests: Write a Python module that removes duplicate objects from a JSON array, keeping the first, with tests. | I need deduplication by a key subset that preserves input order. | Generate a dedupe module and a pytest suite for it.

Produce a standard-library Python module that builds a regular expression from positive and negative examples, plus pytest tests. Choose this when a regular expression must be derived from examples using the shared literal prefix, correctly escaped. Do not choose this expecting a general regex synthesiser: it uses the longest literal prefix shared by the positive examples, and negative examples are used for checking, not for general inference. Example requests: Write a Python module that builds a regex from positive and negative examples, with tests. | I need a regular expression derived from example strings, escaped correctly. | Generate a regex-builder module and prove it with pytest.

Produce a standard-library Python module that turns recorded input/output pairs into a pytest module, plus pytest tests. Choose this when recorded input/output pairs must become one deterministic pytest case per pair. Do not choose this to test behaviour not covered by the recorded pairs, or to generate non-deterministic tests. Example requests: Write a Python module that turns recorded input/output pairs into a pytest module, with tests. | I have input/output examples and want a deterministic regression test suite. | Generate a regression-suite generator and prove it with pytest.

A single go/no-go decision for a release: runs the repository hardening and the supply-chain and licence audit over the same artifacts, combines their findings, and returns a pass or block decision with the blocking findings named. At the end of a release cycle, when a single reproducible pass/block decision is needed. Do not choose this when you need exploitability proven by running the code, or a legal opinion. Example requests: Give me a go/no-go decision for this release. | Is this project safe to ship? Check hardening and licences together. | Which findings block my release? | Run the release readiness gate over these artifacts.

Check a Python requirements file for unpinned packages, duplicate entries (case-insensitively) and pip option lines, with line numbers. Choose this before freezing or auditing a Python requirements file, to find unpinned packages, case-insensitive duplicates and pip option lines. Do not choose this when you need to resolve, install or conflict-check the requirements; it examines the file's text only. Example requests: Which of my Python dependencies are unpinned? | Check this requirements.txt for duplicates before I freeze it. | Is this Python environment file reproducible?

Audit a project's dependency manifest, configuration and source in one pass: leaked secrets, dangerous patterns and unpinned dependencies, with a risk score and grade. Before shipping or reviewing a project, for a fast consolidated security triage. Do not choose this when you need exploitability confirmed by running the code or a CVE database. Example requests: Audit this project for leaked secrets and dangerous patterns. | Give me a security risk score for this code and config. | Check my dependency manifest for unpinned or URL-sourced packages. | Run a security audit pack over these files.

Audit a whole repository's container, infrastructure-as-code and CI artifacts in one pass: privileged containers, wildcard IAM, public buckets, secret interpolation and unpinned build steps, with a weighted risk score, a severity-ranked finding list and a letter grade. Before a release or a security review, to consolidate container, IaC and CI hardening into one ranked report. Do not choose this when you need the image built, the infrastructure applied, or a CVE database consulted. Example requests: Harden my repository before release: check the container, infra and CI. | Find privileged containers and wildcard IAM in these project files. | Give me a risk score for my Dockerfile, Terraform and GitHub workflow. | Run a repository hardening audit over these artifacts.

Build a pain.001 SEPA XML batch from a payments table, validating IBANs and amounts. When a batch of SEPA credit transfers must be produced from a table. Do not choose this when the file must be signed or submitted to a bank. Example requests: Build a SEPA payment file from these rows. | Generate a pain.001 XML batch. | Turn this payment list into a SEPA file.

Analyse a log of x402 payment attempts: settlement rate, every failure reason mapped to its cause, and whether any authorization was left unconsumed. When a payment log needs the failure mix and the settlement rate quantified. Do not choose this when you need on-chain verification of each payment, which needs the chain. Example requests: Why are my x402 payments failing? | Report the settlement rate of these payment attempts. | Which failure reasons dominate my payment log? | Are my payments settling reliably enough to depend on?

Translate a SQL query into a pandas transform module plus a parity test that runs on your sample. When moving a query out of a warehouse into dataframe code. Do not choose this when the query uses joins, CTEs or window functions. Example requests: Convert this SQL query to pandas. | Migrate this query to a dataframe transform. | Give me the pandas equivalent of this SQL.

Split a source schema into fact and dimension tables and generate the star-schema DDL and model notes. When a warehouse needs a dimensional model sketched from sources. Do not choose this when a full normalisation or SCD design is required. Example requests: Design a star schema from this source schema. | Which tables are facts and which are dimensions? | Generate a star schema for this warehouse.

Check a pre-registered analysis plan for completeness and internal consistency before data is collected: the registration identifier, primary outcome, sample size, alpha, power, randomisation, blinding, analysis population, missing-data handling and statistical method, plus uncontrolled multiplicity when several endpoints are named. Before data collection, to catch a plan that is incomplete or internally inconsistent. Do not choose this when you need the power calculation verified against an effect size, which needs the numbers. Example requests: Check my statistical analysis plan before I collect data. | Does my analysis plan state power, alpha and the multiplicity adjustment? | What is missing from this pre-registered analysis plan? | Run a statistical analysis plan check over this protocol.

Compile plain-language strategy rules into a structured rule set and a test module, flagging rules that cannot be compiled. When a strategy described in words must become a structured rule set. Do not choose this when the strategy must be executed or backtested. Example requests: Compile these trading rules into a rule set. | Turn this strategy description into structured rules. | Check whether these rules are implementable.

Assess a project's declared dependency licences and supply-chain hygiene: copyleft and commercially restricted licences, undeclared licences, install-time scripts, URL-sourced packages, leaked credentials in the lockfile and tracked credential files, with a risk score and grade. Before adopting or shipping a dependency set, for a licence-obligation and supply-chain hygiene review. Do not choose this when you need the full resolved dependency tree or a legal opinion on licence compatibility. Example requests: Check my dependencies for copyleft and restricted licences. | Audit my supply chain for install scripts and leaked secrets. | Which of my dependencies have an unknown or undeclared licence? | Run a licence and supply-chain audit over these files.

Review a Terraform module for public ACLs, literal credentials, wildcard IAM policies and disabled encryption, each with the line number. As a fast first pass over a module before a full plan or a policy scan. Do not choose this when you need evaluated plan output or a policy engine such as OPA. Example requests: Review this Terraform module for security problems. | Does this .tf file hardcode any credentials or open a bucket publicly? | Audit my infrastructure-as-code module before I apply it.

Produce a standard-library Python module that parses a text format into a structured result, plus pytest tests including malformed input. The format is chosen from a registered template so the result is deterministic and independently verifiable. Choose this when you need a standard-library parser for a registered text format, delivered with tests that include malformed input. Do not choose this for a bespoke grammar the templates do not cover, or for formats needing a third-party parser library. Example requests: Write a Python module that parses a key=value config file into a dict, with tests for malformed input. | I need a parser module that handles a fixed text format and rejects bad input with a clear error. | Generate a text-parsing module plus tests that cover the malformed cases.

Fit a z-score anomaly threshold to a series and backtest it against the same data. When a series needs a first anomaly threshold and a backtest. Do not choose this when the series has strong seasonality or trend. Example requests: Configure anomaly detection for this series. | What threshold should I use for this time series? | Backtest anomaly detection on this data.

Estimate the input and output token cost of a prompt across a list of model prices. When choosing a model or sizing a prompt budget. Do not choose this when you need exact tokenizer counts, which needs the model's tokenizer. Example requests: Estimate the token cost of this prompt across these models. | What will this prompt cost at this volume? | Compare model prices for this prompt.

Map each tool to the scopes that authorise it and flag tools with no scope. When reviewing which tools an agent may call and under what scope. Do not choose this when you need enforcement, which needs the runtime. Example requests: Build a permission matrix for these tools and scopes. | Which tools have no scope? | Map my agent tools to their permissions.

Reconcile a trade blotter against a broker file, reporting missing trades and value mismatches. At end of day, when the blotter must be tied to the broker's record. Do not choose this when partial fills or allocation differences must be modelled. Example requests: Reconcile my trade blotter against the broker file. | Which trades do not match the broker? | Find the breaks in this trade file.

Produce a standard-library Python module that converts a quantity between named units of length, mass and temperature, plus pytest tests. Choose this when you need to convert a quantity between named units of length, mass or temperature and want the factors exact and the result reproducible, delivered as a module with tests. Do not choose this when you need a unit outside the declared set, currency conversion, or dimensional analysis over compound units such as kilograms per cubic metre: the template covers the units it names and rejects the rest rather than approximating. Example requests: Write a Python module that converts between metric and imperial units with tests. | I need a unit converter for metres, feet, kilograms and pounds that is exact. | Generate a unit-conversion module covering length, mass and temperature and prove it with pytest. | Convert units and measurements: take a quantity in metres, feet, kilograms or pounds and return it in another named unit, with tests.

Check a list of URLs for syntax problems before a crawler, sitemap or link checker uses them: missing scheme, non-numeric port, embedded space, uppercase host. Reports one entry per line with its issues. Choose this when you have a list of URLs and want the syntax problems listed per line — missing scheme, a non-numeric port, an embedded space or a mixed-case host — before a crawler or link checker consumes them. Do not choose this when you need to know whether a URL is reachable: the service makes no network request and cannot report liveness. Example requests: Check this list of URLs for typos before I run the crawler. | Which of these URLs are malformed and why? | Sanity-check my sitemap URLs before I submit them.

Apply jurisdiction tax rates to line items and deliver the tax summary and a line-level audit. When a flat rate must be applied to a set of line items. Do not choose this when reduced rates, thresholds or cross-border rules apply. Example requests: Calculate VAT on these line items. | What is the tax and gross for this basket at the GB rate? | Apply the DE VAT rate to these lines.

Produce a standard-library Python module that validates a webhook payload envelope, plus pytest tests. Choose this when an incoming webhook envelope must be validated to the registered shape and rejected with structured errors. Do not choose this for provider-specific signature verification or for a schema that differs from the registered envelope. Example requests: Write a Python module that validates a webhook JSON envelope and returns structured errors, with tests. | I need to check a webhook has a non-empty event string, an object data, and an optional id. | Generate a webhook-validation module and its pytest suite.