Convert supplied JSON, records and serialization formats, or create bounded artifacts with explicit encodings.
How to choose
Choose by the input format and required output representation. A parser, validator, repairer and canonicalizer have different contracts; inspect accepted schema dialects and numeric semantics.
No file-system access, remote database execution or arbitrary submitted code. Exact integers and decimals require a supported string profile.
WHEN TO USE: Normalize JSON for cache keys or repeatable diffs. Recursively sorts object keys, preserves array order, and emits compact JSON plus UTF-8 byte length. Uses JavaScript numbers and JSON.stringify: integer-index keys serialize numerically before other sorted keys; large integers may already have lost precision. This is not RFC 8785 cryptographic canonicalization. Maximum 100,000 serialized characters, 1,000 entries per collection, depth 32.
Choose for: Sort JSON object keys for reproducible serialization.
Outside this profile: Cryptographic RFC 8785 canonicalization or lossless arbitrary-precision numbers.
WHEN TO USE: Extract one precisely addressed value from an API payload without a query language. Resolves RFC 6901 string pointers, including ~0 and ~1 escaping and the empty root pointer. Missing members, URI fragments, leading-zero array indexes, and reserved prototype-related keys are errors. Maximum document size 100,000 characters and depth 32.
Choose for: Read one value at an exact JSON Pointer.
Outside this profile: JSONPath queries, wildcard selection, or remote API fetching.
WHEN TO USE: Apply a bounded update plan atomically before saving an API document. Implements RFC 6902 add, remove, replace, move, copy, and structural test on a private clone; stops on the first error. Strict operation schemas reject extra fields. Up to 100 operations; size, depth, and 1,000-item limits checked after each operation. Removing the root returns documentExists=false and document=null; adding the root restores it. Reserved prototype-related paths are rejected.
Choose for: Apply an explicit add/remove/replace/move/copy/test plan to supplied JSON.
Outside this profile: Generating a patch from prose or writing to a remote system.
WHEN TO USE: Apply simple partial-object updates using RFC 7396. Object members recursively merge, null members delete, arrays replace wholesale, and non-object patches replace the entire target. The target is never mutated. Maximum 100,000 serialized characters, 1,000 entries per collection, depth 32; reserved prototype-related keys are rejected.
Choose for: Merge partial objects, with null deleting members and arrays replacing wholesale.
Outside this profile: Element-wise array merging or preserving explicit null object members.
WHEN TO USE: Index nested payloads by escaped JSON Pointer for forms or change inspection. Returns scalar leaves in values and every object/array type in containers, preserving empty containers and numeric object keys for lossless unflattening. The root is represented by the empty pointer. Maximum 1,000 total pointers, output 100,000 characters, and depth 32.
Choose for: Produce escaped paths while preserving empty objects and arrays.
Outside this profile: Flattening to ambiguous dotted keys or exceeding 1000 pointers.
WHEN TO USE: Rebuild JSON from json-flatten output without guessing whether numeric keys belong to arrays or objects. Supply scalar values and explicit container types, including the empty root pointer. Rejects leaf/container collisions, missing parents, sparse arrays, unsafe keys, and malformed pointers. Maximum 1,000 combined pointers, 100,000 characters, depth 32.
Choose for: Rebuild a document from explicit scalar leaves and container kinds.
Outside this profile: Guessing missing parent types or accepting sparse arrays.
WHEN TO USE: Shape up to 1,000 records into a small downstream payload, selecting nested fields with JSON Pointers and assigning flat output names. Up to 100 unique output names; missing fields are omitted, set to null, or rejected according to onMissing (default error). Unsafe names are rejected. Maximum 100,000 serialized characters and depth 32.
Choose for: Select JSON Pointer fields from each record and rename them to flat output names.
Outside this profile: Computed expressions, joins, or arbitrary query languages.
WHEN TO USE: Select up to 1,000 records using 1–20 structured predicates with left-to-right short-circuit all/any logic, without executing code. All pointer syntax is validated before filtering. eq/ne use structural JSON equality; gt/gte/lt/lte compare same-type numbers or code-unit strings; contains tests substrings or structural array membership; in tests membership in a supplied list. Missing fields match only exists=false, including for ne. Returns original zero-based indexes. Maximum 100,000 characters and depth 32.
Choose for: Apply structured all/any predicates without code execution.
Outside this profile: Regular expressions, SQL, or fuzzy matching.
WHEN TO USE: Order up to 1,000 records reproducibly using 1–10 nested JSON Pointer keys. Null and missing values tie and follow each key’s nulls setting independently of direction. Remaining ascending type order is boolean, number, string, array, object; booleans use false before true, strings use UTF-16 code units, arrays/objects compare deterministic JSON text. Equal keys preserve input order. Returns original indexes. Maximum 100,000 characters and depth 32.
Choose for: Order records by bounded nested keys with explicit null ordering.
Outside this profile: Locale-aware collation or external database sorting.
WHEN TO USE: Summarize up to 1,000 records with count/sum/avg/min/max, grouped by up to 5 JSON Pointer fields. Up to 20 uniquely named metrics. Group order follows first occurrence; missing keys are distinguished from null through a parallel missing array. Numeric metrics ignore missing/null, reject other non-numbers and overflow; empty sum/count are 0 and empty avg/min/max are null. No groupBy returns one global group, even for empty input. Maximum 100,000 characters and depth 32.
Choose for: Calculate count/sum/avg/min/max over supplied bounded records.
Outside this profile: Database queries, streaming analytics, or arbitrary formulas.
WHEN TO USE: Convert a CSV export into string rows and optional header-keyed records. Handles RFC 4180 quoting, doubled quotes, embedded newlines, and CRLF; also accepts LF/CR line endings, one leading BOM, and optional semicolon/tab/pipe delimiters. Header mode defaults true and requires unique, nonempty, safe headers. Ragged rows and malformed quoting are errors. Values remain strings. Blank lines are one empty field; a final newline adds no row. Maximum 100,000 characters, 1,000 data rows, 100 columns.
Choose for: Parse bounded delimited text with strict quoting and rectangular rows.
Outside this profile: Excel binary files, guessed value types, or malformed CSV repair.
WHEN TO USE: Export scalar rows to CSV with RFC 4180 quoting and spreadsheet formula protection. By default, prefixes an apostrophe to string cells/headers beginning with =,+,-,@, their full-width variants, tab, CR, or LF (also after leading whitespace). Numeric values remain numeric; null becomes an empty cell. escapeFormulas=false is an explicit opt-out for trusted machine-only data; no escaping is universally safe after spreadsheet re-save. Optional unique headers; rectangular rows; CRLF and final newline by default. Maximum 1,000 rows, 100 columns, 100,000 output characters.
Choose for: Serialize scalar rows with default formula-trigger escaping.
Outside this profile: A guarantee of safety after spreadsheet re-save or XLSX generation.
WHEN TO USE: Produce actual inner, left or full joined row pairs from two supplied tables using typed scalar JSON Pointer keys. Enforces declared cardinality over both complete inputs before matching; duplicate null-containing keys count for uniqueness even when nullKeys=never-match. Preserves provenance and order, rejects fanout overruns without partial output. No coercion, fuzzy matching, database access or semantic relationship verification.
Choose for: Join two tables while retaining source rows and indexes Enrich records with a typed lookup and guard duplicate keys Build inner left or full joins with explicit many-to-many fanout
Outside this profile: Fuzzy or semantic entity matching, coercing scalar keys, or proving a business relationship Silently combining colliding columns or truncating a join to fit a cap Fetching data, executing caller code, contacting agents, transmitting a payload, or changing remote state Verifying semantic truth, source authenticity, identity, permissions, or business relationships
WHEN TO USE: Convert supplied localized numeric spellings into exact ASCII decimal strings using a declared digit alphabet, symbols and primary/secondary grouping. Validates the entire spelling with per-item rejections; preserves fractional digit count and negative-zero evidence. No locale inference, floating-point numeric conversion, exponent/currency parsing or full CLDR claim. Maximum 200 items and 128 Unicode code points each.
Choose for: Parse localized decimal strings with an explicit digit and separator profile Convert grouped Arabic-Indic or comma-decimal numbers without floating-point loss Reject partially parsed numbers and retain negative zero and fractional precision evidence
Outside this profile: Inferring a locale, full CLDR/ICU conformance, currency or percentage conversion, exponent notation or accounting parentheses Normalizing Unicode or repairing malformed numeric spellings Fetching data, executing caller code, contacting agents, transmitting a payload, or changing remote state Verifying semantic truth, source authenticity, identity, permissions, or business relationships
WHEN TO USE: Produce contiguous, greedy, byte-bounded JSON-array or LF-terminated NDJSON wire bodies from supplied JSON values. Counts actual UTF-8 bodies including framing and escaping; optional SHA-256 hashes those unchanged bytes. Uses recursive key sorting plus JSON.stringify, not RFC 8785; rejects malformed Unicode and oversized records without partial batches. No transmission, endpoint-limit verification, compression or token accounting.
Choose for: Encode exact byte-limited JSON array or NDJSON request payload batches Pack JSON records into contiguous UTF-8-sized bodies without splitting records Measure escaped JSON framing and hash unchanged batch wire bytes
Outside this profile: Sending requests, validating target endpoint schemas or actual server limits, sizing headers, HTTP framing, compression or tokenization Arbitrary-precision JSON numbers or RFC 8785 cryptographic canonicalization Fetching data, executing caller code, contacting agents, transmitting a payload, or changing remote state Verifying semantic truth, source authenticity, identity, permissions, or business relationships
WHEN TO USE: Rewrite existing supplied JSONC scalar values by guarded RFC 6901 pointers while preserving every source character outside the selected value tokens. Strict object-root JSON plus comments and explicit trailing-comma mode; duplicate decoded keys, malformed untouched regions and unsafe integer-valued numbers fail atomically. No general JSON5, structural edits, configuration security checks or filesystem writes.
Choose for: Change guarded scalar configuration values while preserving JSONC comments and untouched formatting Resolve scalar JSON pointers to source spans before an exact configuration edit
Outside this profile: JSON5 syntax, root arrays, duplicate object names, unsafe integers, scalar type changes, insertions, deletions or structure changes General formatting, schema validation, semantic config safety, arbitrary-precision decimal preservation or file writes Uploading, executing artifacts, contacting agents or registries, publishing, or changing remote state Verifying semantic truth, source authenticity, permissions, target API acceptance or signatures
WHEN TO USE: Construct exact inert multipart/form-data bytes from ordered supplied text or binary file parts, preserving repeated field names. Caller supplies a collision-checked boundary; explicit utf8 or canonical padded Base64 output and decoded wire byte count. Strict ASCII names prevent header injection; no upload, custom headers, nested multipart, automatic field flattening or API acceptance claim.
Choose for: Encode ordered multipart form-data upload bodies with binary files and repeated field names Measure exact multipart request bytes with deterministic collision-checked boundaries
Outside this profile: Performing HTTP requests or uploads, headers beyond Content-Disposition and Content-Type, nested multipart or endpoint compatibility Non-ASCII field names or filenames, path filenames, filename*, or implicit charset guessing Uploading, executing artifacts, contacting agents or registries, publishing, or changing remote state Verifying semantic truth, source authenticity, permissions, target API acceptance or signatures
WHEN TO USE: Build a supplied-input OCI 1.1.1 artifact manifest with SHA-256 digests, exact decoded sizes and inline Base64 bytes for every ordered file layer. Caller declares artifactType and media types; fixed empty JSON config and title-only annotations. No enclosing manifest digest, image build, filesystem layout, registry contact, upload, signature or authenticity guarantee.
Choose for: Compose an OCI 1.1.1 artifact manifest from supplied binary or UTF-8 evidence files Compute content-addressed inline layer descriptors with exact file bytes and source mapping
Outside this profile: OCI image building, registry access, pushing or pulling, tar/layout generation, signatures, provenance truth or publisher authenticity Enclosing manifest digest without exact serialized bytes, inferred timestamps, executable config, platform or subject Uploading, executing artifacts, contacting agents or registries, publishing, or changing remote state Verifying semantic truth, source authenticity, permissions, target API acceptance or signatures
Propose a stable numeric version and escaped release notes from supplied commits under an explicit bounded Conventional Commit footer grammar and pre-1.0 policy.
Choose for: Create release notes and a version proposal from supplied Conventional Commits Apply an explicit breaking-change policy before version 1.0
Outside this profile: Full SemVer/prerelease calculation, all Conventional Commit footer layouts, code-level breaking-change detection, commit-range discovery or revert cancellation Fetching files, reading a repository, executing code or tests, contacting agents, publishing releases, or changing remote state
Find a proved minimum-cost subset covering up to 12 supplied requirement labels, then minimize cardinality and UTF-16 lexicographic IDs across at most 48 declared candidates.
Choose for: Find minimum-cost declared checks covering all supplied requirement labels Choose an exact small weighted set cover with deterministic tie-breaking
Outside this profile: Real test coverage, code impact, reliability, scheduling, inferred capabilities, personnel selection or high-impact decisions about people Large or approximate optimization, negative/zero costs, repeated candidates or extra constraints Fetching files, reading a repository, executing code or tests, contacting agents, publishing releases, or changing remote state
Create exact UTF-8 iCalendar text for up to 64 supplied standalone date-only or UTC events, with caller UIDs/stamps, escaped literal text and byte-bounded folding; no account changes or invitations.
Choose for: Create an ICS file from supplied standalone events Encode all-day or UTC calendar events with exact literal text and caller identifiers
Outside this profile: Calendar booking, finding free time, sending invitations or proving import success Recurrence, attendees, organizers, alarms, URLs, attachments, custom properties, time zones or floating local time Uploading artifacts, executing database statements, sending invitations, accessing calendars or changing remote state
Compile one strict single-table SELECT AST for SQLite3 or PostgreSQL18 into quoted SQL, ordered bindings and source paths, with bounded predicates and explicit pagination; never execute SQL.
Choose for: Compile a bounded SELECT query and parameter bindings from a structured plan Build a single-table SQLite3 or PostgreSQL18 query handoff without interpolating literal values
Outside this profile: Running SQL, database access, schema/type/permission checks or certifying safe execution Raw SQL, writes, joins, wildcard, aliases, expressions, functions, subqueries, arbitrary SQL dialects, floats or inferred types Uploading artifacts, executing database statements, sending invitations, accessing calendars or changing remote state
Expand a strict date-only daily, weekly or monthly rule into a complete bounded window, with COUNT/UNTIL, explicit inclusion/exclusion dates and normalized supported RRULE text; no timezone or account access.
Choose for: Expand supplied date-only recurrence rules into explicit occurrence dates Generate complete daily, weekly or monthly dates inside a bounded window with exclusions
Outside this profile: General RRULE/ICS parsing, time zones, local times, DST, business calendars or finding free appointment slots Unbounded history, unsupported recurrence selectors, holiday guessing or scheduling account changes Uploading artifacts, executing database statements, sending invitations, accessing calendars or changing remote state
Turn 1–24 supplied labeled nonnegative integer values and an explicit domain into an inert accessible horizontal bar-chart SVG, plain-text alternative and exact geometry.
Choose for: Already-computed nonnegative integer rows with a caller-selected zero-based domain
Outside this profile: Aggregating data, inferring domains, negative/fractional values, log scales, stacked bars or time axes Rendering, printing, writing files, uploading, publishing, or changing remote state Supplying custom SVG, HTML, CSS, fonts, images, links, scripts or external resources Guaranteeing font appearance, printer alignment, source authenticity or accessibility compliance
Fit 1–100 short ASCII labels onto supplied integer-micrometre stock, skip first-page used slots and return deterministic row-major placements and up to 20 inert printable SVG sheets.
Choose for: Explicit micrometre page, margin, gap and cell dimensions with one short line per label
Outside this profile: Printer control, barcode generation, postal verification, rotation, optimal nesting, supplied fonts or automatic text fitting Rendering, printing, writing files, uploading, publishing, or changing remote state Supplying custom SVG, HTML, CSS, fonts, images, links, scripts or external resources Guaranteeing font appearance, printer alignment, source authenticity or accessibility compliance
WHEN TO USE: Package 1–32 supplied named UTF-8 or binary artifacts into exact reproducible regular-file USTAR bytes, canonical Base64, SHA-256 and a sorted source-index/offset manifest. Restricted portable ASCII relative paths, fixed nonexecutable metadata, no compression, archive extraction, filesystem reads/writes or universal safe-extraction claim.
Choose for: Package supplied named artifacts as an exact reproducible tar archive Create a byte-addressed bundle with content hashes and source indexes
Outside this profile: ZIP, gzip, compression, directories, links, devices, executable modes, permissions or ownership preservation Reading paths, extracting archives, PAX/GNU extensions or universal cross-platform extraction safety Sending, uploading, extracting archives, writing files, authenticating identities, or changing external state
WHEN TO USE: Create exact inert .eml message text from caller-supplied From, ordered To, UTC timestamp, Message-ID, subject and text. Restricted ASCII mailboxes, deterministic MIME Base64 UTF-8 body and safe folded encoded subjects. Does not send email, create mailbox drafts, authenticate identities, look up domains or promise delivery.
Choose for: Compose a complete deterministic .eml draft artifact for later caller-controlled import or transport Encode explicit sender, recipients, timestamp, message ID and plain text as an RFC MIME message
Outside this profile: Sending email, mailbox access, deliverability checks, DNS or authenticated sender identity HTML, attachments, display names, quoted mailboxes, groups, SMTPUTF8, Cc/Bcc, custom headers or DKIM Sending, uploading, extracting archives, writing files, authenticating identities, or changing external state
WHEN TO USE: Turn an explicit rectangular string matrix into a paste-ready GitHub Flavored Markdown table with exact literal-cell escaping and supplied column alignments. Escapes every ASCII punctuation character, preserving cell text in the declared GFM profile. ASCII @ is rejected to prevent GFM email autolinks. No raw Markdown, HTML, links, newline substitution, display-width padding or non-GFM rendering guarantee.
Choose for: Compose paste-ready literal-cell GFM tables from supplied rows Serialize a rectangular string matrix as Markdown with explicit column alignments
Outside this profile: ASCII @ or email-address cells; GFM email autolinking cannot be suppressed with punctuation escapes Parsing tables, CSV conversion, raw Markdown or HTML cells, links, code spans or multiline cells Pixel alignment, display-width padding or exact behavior on non-GFM renderers; browsers may collapse repeated spaces Sending, uploading, extracting archives, writing files, authenticating identities, or changing external state
Encode 1–64 supplied printable ASCII bytes as a bounded Model 2 QR code using explicit version, error correction and mask; return unbordered module rows or inert black-and-white SVG with a four-module quiet zone.
Choose for: Caller-selected version 1–5, ECC and mask for a short printable ASCII payload
Outside this profile: Opening or judging encoded URLs, verifying destination safety, authenticating payloads or transferring funds Unicode/ECI, arbitrary binary, automatic segmentation/version/ECC, Kanji, micro-QR, logos, colors or camera/print reliability guarantees Rendering, printing, writing files, uploading, publishing, or changing remote state Supplying custom SVG, HTML, CSS, fonts, images, links, scripts or external resources Guaranteeing font appearance, printer alignment, source authenticity or accessibility compliance
WHEN TO USE: Package 1–32 supplied files as exact deterministic ZIP STORE bytes, CRC32, SHA-256 and member offsets for ZIP consumers. Restricted portable ASCII paths, fixed metadata, no compression, extraction, ZIP64, encryption, links or executable permissions. Does not create office-document formats or guarantee safe consumer extraction.
Choose for: Need exact classic ZIP STORE bytes from explicit named file data
Outside this profile: Reading or extracting ZIP, compression, encryption, signing, ZIP64 or filesystem paths Creating DOCX, XLSX or EPUB semantic package formats Extracting archives, writing files, uploading, publishing, or changing external state
WHEN TO USE: Convert caller-supplied Base64 text-file bytes between explicit UTF-8, UTF-16LE, UTF-16BE and reversible Latin1 byte mapping with strict BOM rules. Rejects malformed sequences and unrepresentable scalars without replacements. No charset detection, Windows-1252 aliases, normalization, newline conversion or plaintext echo.
Choose for: Know both source and target encoding and explicit initial-BOM policy
Outside this profile: Guessing encodings, repairing mojibake, lossy replacement, newline conversion or Unicode normalization Windows-1252, ASCII aliases, ISO-8859-1 WHATWG label handling or arbitrary legacy code pages Blind transcoding of fixed byte-column files whose layout depends on original encoding Extracting archives, writing files, uploading, publishing, or changing external state
WHEN TO USE: Remove every PNG chunk of explicitly selected tEXt, zTXt, iTXt, eXIf or tIME types while retaining other bytes exactly. Checks CRC/framing and restricted static 8-bit RGB/RGBA headers only. Does not decode IDAT, guarantee valid images, sanitize all metadata, anonymize pixels or guarantee unchanged visual interpretation; EXIF removal may change orientation.
Choose for: Need bounded chunk-type removal under a container-framing-only guarantee
Outside this profile: Complete metadata scrubbing, privacy sanitization, anonymization, valid-image certification or guaranteed visual equivalence JPEG, WebP, HEIC, TIFF, palettes, transparency keys, APNG, interlacing or unsupported PNG headers Decoding pixels, validating compressed image data, decompressing embedded text or parsing EXIF/ICC semantics Extracting archives, writing files, uploading, publishing, or changing external state
Remove object members or replace selected JSON subtrees with caller-supplied scalars using bounded typed selectors. Original-snapshot matching, expected-count guards and overlap rejection make edits atomic. Optional audit omission; no secret discovery, PII inference, anonymization or safe-to-share guarantee.
Choose for: Caller knows which structural fields must be removed or replaced Guard a repeated-record cleanup with exact match counts and simultaneous edits
Outside this profile: Discover secrets or infer personal information from meaning Guarantee anonymity or that remaining data is safe to share Apply arbitrary JSONPath, regex, recursive descent or scripts Remove array elements or reorder their identity Known exact-pointer edits already fit JSON Patch Read files, send documents, fetch data or change external state
Construct nested destination records from explicit copy, literal and exact-string lookup rules. Return bounded provenance and optional root-test-guarded RFC 6902 patches. Object destinations only; no implicit coercion, expressions, array expansion, writes or source authenticity checks.
Choose for: Map supplied records to nested object destinations using a bounded declarative mapping
Outside this profile: Flat field selection only; use records-project General expressions, recursive array mapping, joins, coercion or redaction Executing code, writing files, environment changes, network requests, publishing or deployment
Compose a local SHA-256 Merkle content commitment binding exact IDs, bytes and item count, plus selected shape-aware inclusion paths. This is consistency with a separately trusted root, not identity, signature, authenticity, confidentiality, completeness, timestamp or file-safety verification.
Choose for: Hand off one supplied artifact and its logarithmic inclusion path without resending the full bundle
Outside this profile: Signatures, certificate transparency, blockchain protocol compatibility, append-only proofs or trusted root distribution Authentication, identity, confidentiality, timestamping, safety or complete-item-list verification Executing code, writing files, environment changes, network requests, publishing or deployment
Compose a deterministic literal .env.example and unresolved-variable checklist from explicit public and names-only secret declarations. Secret declarations accept no value/default field. Names-only secret declarations; public values supplied by the caller. No secret detection, secret-free certification, environment reads or changes.
Choose for: Compile explicit declarations into a deterministic dotenv example with literal public defaults
Outside this profile: Secret scanning, secret-free certification, retrieving credentials or classifying whether caller public values contain secrets Shell sourcing, escape syntax, variable interpolation, production configuration or environment mutation Executing code, writing files, environment changes, network requests, publishing or deployment
Compose all grid-aligned candidate meetings with whole-slot attendee continuity, required participants and explicit meeting buffers from supplied available/busy intervals.
Choose for: find candidate meeting slots from supplied calendars enumerate whole-meeting attendee availability compose buffered quorum meeting candidates
Outside this profile: Live account access, invitations, booking or authenticating external observations Aggregate concurrency without the same attendees available for the entire meeting Ranked preferences, scheduling optimization or a conflict-free set of meetings Accessing calendars, scheduling jobs, reserving meetings, network requests or external state changes
Compose an aligned grid from supplied nonoverlapping piecewise-constant segments with exact rational integrals, means over supplied support and source contribution durations.
Choose for: resample piecewise-constant time series to an exact aligned grid calculate time-weighted means over explicit supplied support compose exact rational series cells with missing-support evidence
Outside this profile: Interpolation, zero filling, extrapolation or last-observation carry-forward Overlapping segments within one series or undeclared measurement validity Floating-point approximations, display rounding or inference of observation quality Accessing calendars, scheduling jobs, reserving meetings, network requests or external state changes
Expand bounded five-field cron into complete UTC occurrences using an exact caller-supplied offset timeline, including every fold instant and no nonexistent gap minute.
Choose for: expand cron occurrences in a supplied UTC window enumerate fixed-offset or UTC cron schedules offline inspect cron gap and fold matches against supplied offset periods
Outside this profile: An IANA timezone name without an explicit trusted offset-period timeline Daemon scheduling, Cronie clock-jump replay, catch-up or once-per-wall-time suppression Quartz, seconds/year fields, names, macros, hashed values, command suffixes or Sunday 7 Accessing calendars, scheduling jobs, reserving meetings, network requests or external state changes
Compose actual proto3 wire bytes, matching explicit schema and exact field offsets from bounded acyclic integer/string/bytes declarations. Deterministic under this profile only; no universal canonical serialization, RPC, transport, credentials or execution.
Choose for: Prepare an inert request body or artifact for a separately authorized Protobuf consumer
Outside this profile: Recursive/imported schemas, floats, fixed-width types, enums, maps, oneofs, extensions or well-known types gRPC framing, sending a request, signatures or cross-implementation byte canonicality Executing code, writing files, environment changes, network requests, publishing or deployment
Compile two to four supplied deterministic finite event constraints into a minimized synchronous intersection machine, exact member classes and a shortest accepted trace. Missing transitions reject; no asynchronous interleaving, runtime execution, liveness, deadlock or authorization guarantee.
Choose for: Every component intentionally consumes every event from the same explicit common alphabet
Outside this profile: Asynchronous components, implicit unrelated-event self-loops, epsilon edges, guards, actions, counters or real-time predicates Policy appropriateness, safe concurrency, liveness, real-world deadlocks or permission to execute actions Executing code, writing files, environment changes, network requests, publishing or deployment
Generate concrete rows covering every pair with a feasible complete assignment under bounded forbidden tuples, with exhaustive-coverage certificate and deterministic greedy ties. No minimum-suite or target-behavior claim.
Outside this profile: Minimum test suites, arbitrary SAT predicates, unlimited Cartesian products or bug-finding guarantees Execute target code, call APIs, fetch data, run external agents, mutate accounts or deploy
Generate reproducible JSON records from explicit integer sequences, uniform integers, Bernoulli draws and integer-weighted categories using sha256-counter-fields-v1. No real-PII source, population realism, privacy certification or cryptographic-randomness claim.
Choose for: Small inert fixture tables from elementary explicit sampling rules
Outside this profile: Exact quotas, realistic identities, PII anonymization, correlated populations, Gaussian or floating-point distributions Cryptographic tokens, secrets or unpredictable random draws Execute target code, call APIs, fetch data, run external agents, mutate accounts or deploy
Select the smallest target-failing subsequence among supplied observations, reconcile conflicting evidence, and materialize missing one-deletion experiments. References are caller assertions; no target execution, reproduced bug or global minimum claim.
Choose for: Caller already owns a harness and supplies bounded pass/fail/unresolved observations
Outside this profile: Autonomous ddmin, reproducing bugs, root-cause claims, global minimality or executing a target Unobserved harness identity, flakiness or causality verification Execute target code, call APIs, fetch data, run external agents, mutate accounts or deploy
Encode typed values or decode a strictly admitted core-deterministic CBOR integer subset with bounded tokens and exact integer precision.
Choose for: A declared consumer requires this specific binary format and exact integer/binary distinctions Create a bounded offline configuration or protocol fixture from supplied typed values
Outside this profile: Floating point, dates, timestamp/tag extensions, missing-value primitives, custom objects, streaming or arbitrary maps General CBOR, tags, indefinite lengths, unsorted keys or nonpreferred encodings Legacy length-first CBOR map ordering or dCBOR normalization Network transport, authentication, signatures, COSE/CWT/WebAuthn, consumer schema validation, or execution of decoded content
Compose stable MessagePack bytes from typed integers, text, binary, arrays and restricted-key maps; exact uint64/int64, encode only.
Choose for: A declared consumer requires this specific binary format and exact integer/binary distinctions Create a bounded offline configuration or protocol fixture from supplied typed values
Outside this profile: Floating point, dates, timestamp/tag extensions, missing-value primitives, custom objects, streaming or arbitrary maps Decode or inspect MessagePack bytes; decoding is explicitly held A universal MessagePack canonical form Network transport, authentication, signatures, COSE/CWT/WebAuthn, consumer schema validation, or execution of decoded content
Compose deterministic TOML 1.0-compatible configuration from explicit string, signed-int64, boolean and homogeneous-array declarations; preserve integer precision and detect table conflicts.
Choose for: Creating new TOML 1.0-compatible files from explicit typed declarations Preserving full signed-int64 precision without JavaScript numeric conversion Producing an inert exact-byte configuration artifact for an offline consumer
Outside this profile: Parsing, merging or editing existing TOML source or preserving comments Floating-point values, dates, times, null, nested or mixed arrays, arrays of tables or inline tables Application configuration validation, interpolation, environment expansion, filesystem includes or secret classification Network requests, writing files, live configuration changes, authentication, publishing or deployment
Compose separate exact RFC 8941-compatible structured field values from explicit typed Items, Lists and Dictionaries; preserve decimal types without floating-point rounding.
Choose for: Composing caller-declared structured field name/value pairs for offline protocol fixtures Preserving ordered dictionaries and parameters with explicit string/token/bytes/integer/decimal/boolean types Rejecting decimal rounding, injection and duplicate keys before handoff
Outside this profile: Parsing existing fields, combining repeated lines or normalizing arbitrary headers Full RFC 9651 Date and Display String types; this explicit RFC 8941 vocabulary profile excludes them HTTP Message Signatures, signing, digest verification, authentication, cookies, header policy or application field semantics Raw HTTP header blocks, requests or transmission Network requests, writing files, live configuration changes, authentication, publishing or deployment
Build an optional, open observational JSON Schema draft and counted evidence from supplied JSON samples; never infer a real API contract.
Choose for: Review a recursive open schema draft inferred from bounded supplied examples Inspect sample counts, optional properties and unobserved empty-array items
Outside this profile: Establishing a universal API contract or production rejection policy Inferring required fields, enums, formats, ranges, cross-field correlations or compatibility Generating a tool definition or assuming generalized fixture-generator compatibility Network access, filesystem changes, target execution, publishing or deployment
Factor supplied JSON log records into exact literal templates and typed caller-selected scalar occurrences, preserving parsed values and record order for client reconstruction.
Choose for: Preserve every parsed JSON log value in exact literal templates and ordered occurrences Select scalar parameter pointers explicitly and reconstruct original records on the client
Outside this profile: Semantic log grouping, anomaly detection, root cause, severity or incident diagnosis Automatic timestamp, identifier, UUID, IP or secret detection and redaction Guaranteed compression, lower model token cost, authenticated history or safer sharing Preserving source member order, duplicate keys, numeric token spelling or pre-parse precision Network access, filesystem changes, target execution, publishing or deployment
Canonicalize strict supplied JSON source with duplicate-name, Unicode and binary64 admission to bounded RFC 8785-compatible UTF-8 text and SHA-256. No signing, authenticity, authorization or lossless decimal guarantee.
Choose for: An external consumer explicitly requires the bounded jcs-rfc8785-source-binary64-v1 profile Source JSON must retain duplicate-member visibility and recursively sorted UTF-16 object keys
Outside this profile: Lossless arbitrary-precision decimal or large-integer semantics; represent such data as strings Repairing JSON, authorizing actions, signing, authenticating authors or proving content correctness Secrets or personal data that must not be disclosed; unsalted artifact hashes do not anonymize inputs