SwishrJobs

Produce a standard-library Python module that reports boundary conditions for a text value, plus pytest tests. Choose this when a text value must be classified against a fixed set of boundary conditions, one boolean each. Do not choose this for validation with error messages (use data-validation-module) or for conditions outside the fixed set. Example requests: Write a Python module that reports which boundary conditions a text value satisfies, with tests. | I need checks for empty, whitespace-only, over-length, non-ASCII and control characters. | Generate an edge-case check module and prove it with pytest.

Compare two sets of texts and report the closest match and any near-duplicates. When deduplicating or matching records by text similarity. Do not choose this when semantic similarity is required across different vocabulary. Example requests: Find near-duplicate texts between these two sets. | Compare these two sets of texts for overlap. | Which of these records are duplicates of each other?

Produce a standard-library Python module that parses a dotenv-style configuration into JSON, plus pytest tests. Choose this when you need to parse a dotenv-style configuration into a JSON object, skipping comments and blank lines and honouring quoted values, delivered as a module with tests. Do not choose this when you need variable expansion, multi-line values, or to read the real process environment: this parser takes a string and returns an object, with no interpolation. Example requests: Write a Python module that parses a .env file into a dict, handling comments and quotes, with tests. | I need a dotenv parser that ignores blank lines and export prefixes. | Generate a dotenv-parsing module and prove it with pytest.

Check a .env file for duplicate keys, empty values, unquoted values with spaces, invalid names and values that look like leaked credentials. Reports each problem with its line number. Choose this before a deploy, or when auditing an environment file, to find duplicate keys, empty values, unquoted values with spaces, invalid names and values whose shape looks like a credential. Do not choose this for a secrets manager or vault, which it cannot read, and do not treat a clean report as proof that no secret is present: the detector is a shape heuristic. Example requests: Check my .env for duplicate keys before I deploy. | Did I leave an empty value or a leaked secret in this environment file? | Audit this dotenv file for mistakes.

Turn a table of task examples into a JSONL evaluation set and a scoring rubric. When you have labelled examples and need an eval set and rubric. Do not choose this when you need synthetic examples generated, which needs a model. Example requests: Build an eval dataset from these examples. | Turn this table of task examples into a JSONL eval set. | Make an evaluation set and rubric for this task.

Check an expense claim against an amount limit, allowed categories and receipt rules. When a claim must be checked against a stated policy. Do not choose this when the policy varies by role, project or period. Example requests: Check these expenses against our policy. | Which expense claims break the rules? | Audit this expense claim for missing receipts.

Find feature flags that are declared but appear nowhere in the source, and deliver a removal plan. During flag cleanup, when you have the flag list and the source. Do not choose this when flags are resolved dynamically, which a textual check cannot follow. Example requests: Which feature flags can I delete? | Find dead feature flags in this source tree. | Audit my flags and tell me what is safe to remove.

Produce a standard-library Python module that renames and selects JSON fields using a mapping, plus pytest tests. Choose this when fields must be renamed and selected via an explicit source-to-destination mapping. Do not choose this for computed or transformed values: only names and selection change. Example requests: Write a Python module that renames and selects JSON fields using a mapping, with tests. | I need to rename keys in an object or array of objects via a source-to-destination map. | Generate a field-mapping module and test it.

Compute standard financial ratios from a trial balance and report any missing inputs. When a trial balance or account summary needs standard ratios. Do not choose this when account names must be mapped from a chart of accounts. Example requests: Compute the financial ratios from this trial balance. | What are my margins and current ratio? | Give me a ratio report from these figures.

Convert raw instruction/response pairs into training JSONL in chat or completion format. When preparing labelled pairs for a fine-tuning run. Do not choose this when you need data quality judged, which needs a model. Example requests: Format these pairs for fine-tuning. | Convert my instruction/response data to training JSONL. | Prepare this dataset in chat format.

Convert a column of amounts to a base currency using a supplied rate table. When amounts in mixed currencies must be brought to one base. Do not choose this when live or date-specific rates are required. Example requests: Convert these amounts to USD. | Normalise this column to a single currency. | Apply this rate table to my rows.

Turn a list of guardrail rules into a policy document and a test module, and check examples against it. When codifying content or safety rules for an agent. Do not choose this when you need a live moderation model, which needs a model call. Example requests: Build a guardrail policy from these rules. | Generate a policy config and tests for these rules. | Check these examples against my guardrails.

Produce a standard-library Python module that extracts readable text from an HTML fragment, plus pytest tests. Choose this when you need the visible text of an HTML fragment with tags removed, script and style contents dropped, entities unescaped and whitespace collapsed, delivered as a module with tests. Do not choose this when you need a parsed DOM, CSS-selector queries, or a faithful browser rendering: the result is a text approximation, not a parsed document. Use a full HTML parser for that. Example requests: Write a Python module that strips HTML tags from a fragment and returns the plain text, with tests. | I need to turn an HTML snippet into readable text, dropping script and style blocks. | Generate an HTML-to-text module and prove it with pytest.

Validate IBAN checksums and BIC formats across a table of accounts. Before a payment run, to catch malformed account identifiers. Do not choose this when account existence must be confirmed, which needs the bank. Example requests: Validate these IBANs and BICs. | Which of these account numbers are wrong? | Check the checksums on this IBAN list.

Extract invoice number, date, total and VAT from invoice text and deliver a reconciliation CSV. When invoices arrive as text and need structuring for reconciliation. Do not choose this when the source is a PDF or scan, which needs OCR. Example requests: Extract the totals from these invoices. | Turn this invoice text into structured records. | Parse these invoices into a reconciliation file.

Validate an ISO 20022 payment message for its required blocks and fields, and total the amounts. Before sending a payment message, to catch missing elements. Do not choose this when full XSD schema validation is required. Example requests: Validate this ISO 20022 payment message. | Is my pain.001 message complete? | Check this payment XML for missing fields.

Authorize and execute a previously quoted job, then return the verified result. Requires explicit authorization. Prefer a capability-specific tool (for example submit_csv_json_transformation) when one exists, because it documents that capability's exact inputs.

Produce a standard-library Python module that renders a JSON array of objects as CSV text, plus pytest tests. Choose this when you need a tested function that renders a JSON array of objects as CSV, using the union of keys as the header. Do not choose this when you need the conversion performed on your data now, or when the CSV format must be bespoke (delimiter, quoting rules). Example requests: Write a Python function that renders a JSON array of objects as CSV, with tests. | I need JSON to CSV with the header being the union of all keys. | Generate a JSON-to-CSV module and prove it with pytest.

Produce a standard-library Python module that reports the structural difference between two JSON documents, plus pytest tests. Choose this when you need the structural difference between two JSON documents as added, removed and changed paths, delivered as a module with tests. Do not choose this for a textual or line-oriented diff, for JSON Patch generation, or for applying a patch. This reports differences; it does not transform documents. Example requests: Write a Python module that diffs two JSON documents and reports added, removed and changed keys, with tests. | I need a structural JSON diff that shows changed paths, not a line diff. | Generate a JSON-diff module and prove it with pytest.

Audit the keys in a JSON document: keys that differ only by case, keys containing non-identifier characters, total key count, occurrence frequency and maximum nesting depth. Choose this before generating a schema or code from a JSON sample, to find keys that differ only by case, keys with non-identifier characters, and the document's nesting depth. Do not choose this when you need the document validated against a schema; use json-schema-validation. Example requests: Are the keys in this JSON consistent? | Audit this JSON for casing mistakes before I generate a schema. | How deep is this JSON and what keys does it use?

Validate a JSON Lines (NDJSON) stream and report the line number and parser error of every line that is not valid JSON, so a broken export can be fixed without bisecting it. Choose this when a JSON Lines or NDJSON file fails to load and you need the exact line number and parser error rather than bisecting the file. Do not choose this for a single multi-line JSON document; use json-schema-validation or json-key-audit. Example requests: Find the broken line in this JSON Lines export. | Which lines of this NDJSON file are invalid? | Validate this jsonl stream before I load it.

Produce a standard-library Python module that infers a JSON Schema from sample records, plus pytest tests. Choose this when you need a draft 2020-12 JSON Schema inferred from sample objects, with keys present in every record marked required. Do not choose this for a hand-tuned schema or for constraints the samples do not express (formats, ranges, enums). Example requests: Write a Python module that infers a JSON Schema from sample records, with tests. | I need a JSON Schema generated from example objects, marking common keys as required. | Generate a schema-inference module and prove it with pytest.

Validate a JSON document against a JSON Schema and report every violation with its JSON Pointer path and the keyword that failed. Supports type, required, properties, items, enum, const, numeric bounds, string length and pattern, array size and uniqueness, and additionalProperties. Choose this when you have a concrete JSON document and a JSON Schema and need every violation located by JSON Pointer path and failing keyword. Do not choose this when you need a validator module for your own service (use webhook-payload-validation), or when the schema relies on $ref, allOf, anyOf or oneOf, which this documented subset does not evaluate. Example requests: Validate this JSON payload against my schema and list the failures. | Which fields in this webhook payload violate the schema? | Check this API response against its JSON Schema.

Convert a JSON document into a CSV file. Accepts an array of objects, a single object, or an API envelope such as {"data": [...]}. The header is the sorted union of every record's keys, nested values are kept as compact JSON, and records with missing keys are padded and reported. Choose this when you have JSON records and need a CSV file: an array of objects, a single object, or an API envelope such as {"data": [...]}. The header is the sorted union of every record's keys, so a record with an extra field still contributes a column. Do not choose this when you need the opposite direction (use csv-to-json-rows), when the JSON must be validated against a schema (use json-schema-validation), or when the records need cleaning first (run csv-clean on the output). Example requests: Turn this API response into a CSV I can open in Excel. | Convert this JSON export into CSV for a bulk database import. | Flatten this list of JSON records into a spreadsheet.

Produce a standard-library Python module that rewrites JSON object keys to lower snake_case, plus pytest tests. Choose this when object keys must become lower snake_case, with values left untouched. Do not choose this expecting acronym splitting: a capital run like HTTPServer becomes httpserver, not http_server. Example requests: Write a Python module that rewrites JSON object keys to lower snake_case, with tests. | I need camelCase keys converted to snake_case without changing the values. | Generate a key-normalisation module and a pytest suite.

Turn a list of metric names and source tables into a metrics.yml catalogue with grains and dimensions. When standing up a metrics layer from a list of agreed KPI names. Do not choose this when the metric SQL must be fully implemented and validated. Example requests: Build a metrics catalogue from these KPI names. | Define these metrics against my source tables. | Create a metrics.yml for my KPIs.

Check Kubernetes manifests for privileged containers, root users, host networking, mutable :latest tags and writable root filesystems, and deliver a hardened copy of each document. Before applying manifests, to catch the common pod-security mistakes. Do not choose this when you need admission-control policy evaluation or schema validation. Example requests: Harden this Kubernetes manifest and show me what was unsafe. | Are these deployment manifests running privileged containers? | Give me a hardened version of my k8s YAML.

Check that every journal entry and the ledger as a whole balance, and flag malformed rows. After posting a journal, before closing the period. Do not choose this when account-code or period validation is required. Example requests: Check whether this ledger balances. | Find the unbalanced journal entries. | Audit this journal for double-sided rows.

Validate a JSON model output against a schema, locate every violation, and coerce the fields that can be repaired. After an LLM call, before trusting or storing its structured output. Do not choose this when you need full JSON Schema compliance including combinators. Example requests: Check this model output against my schema. | Guard my LLM JSON output and fix what can be fixed. | Validate this structured output and repair it.

Infer a log format from sample lines and deliver a Python parser module that extracts timestamps, addresses and bracketed fields, plus the parsed samples as JSON. When you have sample logs and want a working parser rather than a regex to write by hand. Do not choose this when the log format is already known and a library parser exists. Example requests: Write me a parser for this log format. | Generate a Python module that parses these log lines. | What fields can be extracted from these logs?

Produce a standard-library Python module that reports records violating a fixed field contract, plus pytest tests. Choose this when you need one entry per offending record, naming its index and the rule it broke, against a fixed contract. Do not choose this for a contract outside the registered template, or if you need the records corrected rather than reported. Example requests: Write a Python module that finds records violating a fixed field contract, with tests. | I need a checker that reports the index and failed rule for each bad record. | Generate a malformed-record detector and prove it with pytest.

Produce a standard-library Python module that renders JSON records as a Markdown table, plus pytest tests. Choose this when JSON records must become a GitHub-flavoured Markdown table with pipes escaped. Do not choose this for HTML tables, spreadsheet formats, or column alignment rules beyond the template. Example requests: Write a Python module that renders JSON records as a GitHub Markdown table, with tests. | I need a Markdown table renderer that escapes pipes so data cannot break the table. | Generate a Markdown-table module and test it.

Lint MCP tool definitions: missing input schemas, non-object schemas, and untyped properties. Before publishing an MCP server, to catch schema defects. Do not choose this when you need the tools executed. Example requests: Lint these MCP tool schemas. | Are my MCP tool definitions well formed? | Check my MCP tools for schema problems.

Produce a standard-library Python module that extracts key/value metadata from a document head, plus pytest tests. Choose this when 'key: value' metadata must be read from the head of a text document, preserving duplicate keys rather than overwriting. Do not choose this for YAML/TOML front matter parsing or for metadata spread through the body. Example requests: Write a Python module that extracts 'key: value' metadata from a document head, with tests. | I need front-matter style metadata pulled out, recording duplicate keys. | Generate a metadata-extraction module and test it.

Assemble a model card from a model config and its metrics, and flag missing governance sections. When a model needs documentation for governance or release. Do not choose this when you need the model evaluated, which needs the model and data. Example requests: Generate a model card from this config and these metrics. | Document this model for governance review. | Build a model card for my classifier.

Split a workflow description into typed steps and produce an orchestration plan and a Mermaid diagram. At design time, to turn a described workflow into steps and a diagram. Do not choose this when you need the workflow executed, which needs the agents. Example requests: Plan the orchestration for this agent workflow. | Turn this workflow into agent steps and a diagram. | Design the multi-agent flow for this process.

Produce a standard-library Python module that flattens a nested object into dotted-path keys, plus pytest tests. Choose this when a nested object must become single-level dotted-path keys, leaving lists as leaves. Do not choose this if you need lists indexed into keys (a.0, a.1) or a custom separator: the template fixes the behaviour. Example requests: Write a Python module that flattens a nested object into dotted-path keys, with tests. | I need nested JSON flattened to single-level keys like a.b.c. | Generate a flattening module and prove it with pytest.