TypeScript / Execução registrada
agent-work-layer
Work Orders and Work Reports for AI agents. A typed SDK, validator, lifecycle engine and MCP server for handing work from one agent to another, with checkable results.
O que foi verificado
Os comandos abaixo executam o código do repositório e conferem os cenários registrados. A evidência cobre esses cenários na versão indicada; não comprova todas as entradas possíveis nem uso em produção.
Este repositório é um fork de probita-ai/agent-work-layer. A execução corresponde ao fork; a autoria e a licença original são preservadas.
A evidência abaixo pertence à versão registrada. A consulta usa a API pública do GitHub e pode atingir o limite de requisições.
Comandos e resultados reais
Resultados extraídos dos logs do Actions. Saídas extensas podem estar resumidas; a execução original e o artefato contêm o registro completo.
Baixar relatório JSONPASS npm ci
Saída: 0 · esperada: 0 · 2.106 s
$ bash -e -o pipefail -c npm ci added 99 packages, and audited 100 packages in 2s 34 packages are looking for funding run `npm fund` for details found 0 vulnerabilities
PASS npm run typecheck
Saída: 0 · esperada: 0 · 0.402 s
$ bash -e -o pipefail -c npm run typecheck > agent-work-layer@0.1.2 typecheck > tsc --noEmit -p tsconfig.json
PASS npm test
Saída: 0 · esperada: 0 · 7.694 s
$ bash -e -o pipefail -c npm test > agent-work-layer@0.1.2 test > node --test "test/**/*.test.ts" ▶ awl CLI ✔ --help and no arguments print usage (1084.845887ms) ✔ --version prints the package version (513.752987ms) ✔ validate passes the good example (347.668409ms) ✔ validate fails the bad example and names each rule (346.427734ms) ✔ validate --json prints machine-readable results (346.247125ms) ✔ validate skips the report when the order is invalid (303.47964ms) ✔ validate --parent checks delegation rules (656.541635ms) ✔ a warning-only report exits 0 and marks the file with ! (304.321629ms) ✔ usage errors exit 2 (1528.40313ms) ✔ schema prints both JSON Schemas (615.239528ms) ✔ awl CLI (6049.21942ms) ▶ awl desk over stdio ✔ two processes coordinate through a shared folder (895.889122ms) ✔ awl desk over stdio (908.850793ms) ▶ toolAllowed ✔ filings/search with [filings/search] → true (1.395997ms) ✔ filings/search with [filings/*] → true (0.344817ms) ✔ filings/search with [*] → true (0.190267ms) ✔ filings/search with [filings/get] → false (0.310032ms) ✔ filings/search with [filingsx/*] → false (0.201608ms) ✔ filings/search with [files/*, calc/*] → false (0.367619ms) ✔ filings/search with [] → false (0.277991ms) ✔ toolAllowed (5.34424ms) ▶ patternCovered ✔ filings/search within [filings/search] → true (1.439519ms) ✔ filings/search within [filings/*] → true (0.468739ms) ✔ filings/* within [filings/*] → true (0.401193ms) ✔ filings/* within [*] → true (0.338765ms) ✔ * within [*] → true (0.099347ms) ✔ filings/* within [filings/search] → false (0.620464ms) ✔ * within [filings/*] → false (0.113352ms) ✔ calc/sum within [filings/*] → false (0.109285ms) ✔ calc/sum within [] → false (0.315962ms) ✔ patternCovered (5.01348ms) ▶ validateChildOrder ✔ accepts a strictly narrower child (3.374807ms) ✔ accepts a child equal to its parent (0.192821ms) ✔ accepts extra budget units the parent does not limit (0.207209ms) ✔ C1 rejects a child that drops a parent budget unit (0.473928ms) ✔ C1 rejects a larger budget (0.193843ms) ✔ C2 rejects a wider tool pattern (0.383208ms) ✔ C2 rejects a tool the parent does not have (0.170239ms) ✔ C3 rejects a later deadline (0.168026ms) ✔ reports every violation at once (0.443211ms) ✔ validateChildOrder (6.137908ms) ▶ create ✔ fills spec, id, created_at and from (7.438266ms) ✔ keeps an explicit id (0.331251ms) ✔ the actor is always the manager, whatever the input says (0.394199ms) ✔ uses a random id by default (2.423584ms) ✔ rejects a duplicate id (1.373174ms) ✔ rejects an invalid order with findings (0.443551ms) ✔ rejects a deadline already in the past (0.348424ms) ✔ requires an actor (1.262878ms) ✔ create (16.488935ms) ▶ reads ✔ get throws NOT_FOUND for an unknown id (0.888195ms) ✔ returned records are copies (0.628218ms) ✔ list filters by to, from, state and parent (17.010172ms) ✔ list returns summaries (0.509956ms) ✔ reads (19.514869ms) ▶ lifecycle through the desk ✔ happy path: offer, accept, start, report, approve (3.746364ms) ✔ the report gets spec, order_id, from, attempt and finished_at (0.99747ms) ✔ a worker-supplied finished_at is kept (0.827841ms) ✔ decline ends the order (0.941695ms) ✔ rework returns to working and numbers attempts (1.537593ms) ✔ rework needs a reason (1.328591ms) ✔ review rejects an unknown decision (0.703138ms) ✔ blocked, answered, then done (1.894883ms) ✔ an empty answer is refused (0.658224ms) ✔ a failed report Full transcript in the execution-proof artifact.
PASS npm run test:package
Saída: 0 · esperada: 0 · 3.964 s
$ bash -e -o pipefail -c npm run test:package > agent-work-layer@0.1.2 test:package > npm run build && node scripts/package-smoke.ts > agent-work-layer@0.1.2 build > rm -rf dist && tsc -p tsconfig.build.json && chmod +x dist/cli.js package ok: agent-work-layer-0.1.2.tgz, 27 files, 26.5 kB
Reproduzir a execução
Os pré-requisitos e as entradas são públicos no repositório. Faça checkout da versão exata antes de executar os cenários.
git clone https://github.com/brunnojob/agent-work-layer.git cd agent-work-layer git checkout 345b0163da5819b3af6bf58aa09d74e84ab7e4fd python .proof/record.py