Skip to content

Skill harness changes part 5#3224

Merged
vigoo merged 34 commits intomainfrom
skill-harness-5
Apr 22, 2026
Merged

Skill harness changes part 5#3224
vigoo merged 34 commits intomainfrom
skill-harness-5

Conversation

@vigoo
Copy link
Copy Markdown
Contributor

@vigoo vigoo commented Apr 20, 2026

New Skills (88 new SKILL.md files)

Common (language-independent) — 17 new

  • golem-add-component
  • golem-cancel-queued-invocation
  • golem-cloud-account-setup
  • golem-configure-mcp-server
  • golem-delete-agent
  • golem-edit-manifest
  • golem-interrupt-resume-agent
  • golem-list-and-filter-agents
  • golem-local-dev-server
  • golem-manage-plugins
  • golem-profiles-and-environments
  • golem-redeploy-agents
  • golem-rollback
  • golem-test-crash-recovery
  • golem-troubleshoot-build
  • golem-update-running-agents
  • golem-view-agent-files

MoonBit — 35 new

  • golem-add-agent-moonbit
  • golem-add-config-moonbit
  • golem-add-cors-moonbit
  • golem-add-http-auth-moonbit
  • golem-add-http-endpoint-moonbit
  • golem-add-ignite-moonbit
  • golem-add-llm-moonbit
  • golem-add-moonbit-package
  • golem-add-mysql-moonbit
  • golem-add-postgres-moonbit
  • golem-add-secret-moonbit
  • golem-add-transactions-moonbit
  • golem-add-webhook-moonbit
  • golem-annotate-agent-moonbit
  • golem-atomic-block-moonbit
  • golem-call-another-agent-moonbit
  • golem-call-from-external-moonbit
  • golem-configure-durability-moonbit
  • golem-create-agent-instance-moonbit
  • golem-custom-snapshot-moonbit
  • golem-file-io-moonbit
  • golem-fire-and-forget-moonbit
  • golem-http-params-moonbit
  • golem-interactive-repl-moonbit
  • golem-invoke-agent-moonbit
  • golem-make-http-request-moonbit
  • golem-multi-instance-agent-moonbit
  • golem-parallel-workers-moonbit
  • golem-recurring-task-moonbit
  • golem-schedule-agent-moonbit
  • golem-schedule-future-call-moonbit
  • golem-stateless-agent-moonbit
  • golem-trigger-agent-moonbit
  • golem-wait-for-external-input-moonbit

Rust — 12 new

  • golem-add-ignite-rust
  • golem-add-llm-rust
  • golem-add-mysql-rust
  • golem-add-postgres-rust
  • golem-add-webhook-rust
  • golem-call-from-external-rust
  • golem-create-agent-instance-rust
  • golem-custom-snapshot-rust
  • golem-interactive-repl-rust
  • golem-parallel-workers-rust
  • golem-recurring-task-rust
  • golem-wait-for-external-input-rust

Scala — 12 new

  • golem-add-ignite-scala
  • golem-add-llm-scala
  • golem-add-mysql-scala
  • golem-add-postgres-scala
  • golem-add-webhook-scala
  • golem-call-from-external-scala
  • golem-create-agent-instance-scala
  • golem-custom-snapshot-scala
  • golem-interactive-repl-scala
  • golem-parallel-workers-scala
  • golem-recurring-task-scala
  • golem-wait-for-external-input-scala

TypeScript — 12 new

  • golem-add-ignite-ts
  • golem-add-llm-ts
  • golem-add-mysql-ts
  • golem-add-postgres-ts
  • golem-add-webhook-ts
  • golem-call-from-external-ts
  • golem-create-agent-instance-ts
  • golem-custom-snapshot-ts
  • golem-interactive-repl-ts
  • golem-parallel-workers-ts
  • golem-recurring-task-ts
  • golem-wait-for-external-input-ts

.agents/ top-level skills — 5 new

  • moonbit-agent-guide
  • moonbit-c-binding
  • moonbit-code-transform
  • moonbit-proof
  • moonbit-refactoring

Modified Skills (7)

  • golem-get-agent-metadata (common)
  • golem-http-params-rust
  • golem-http-params-scala
  • golem-http-params-ts
  • golem-schedule-future-call-scala
  • golem-schedule-future-call-ts
  • .agents/skills/golem-skill-harness/SKILL.md

New Scenarios — 34

  • add-component
  • add-ignite-rdbms
  • add-llm
  • add-mysql-rdbms
  • add-postgres-rdbms
  • add-webhook
  • auto-update-agent
  • call-from-external
  • cancel-queued-invocation
  • configure-mcp-server
  • create-agent-instance
  • delete-agent
  • edit-manifest-add-environment
  • edit-manifest-add-plugin
  • edit-manifest-presets-and-env-cascade
  • interactive-repl
  • interrupt-resume-agent
  • list-and-filter-agents
  • local-dev-server
  • manage-plugins-cli
  • manage-plugins-manifest
  • manual-snapshot-update-agent
  • parallel-workers
  • profiles-and-environments-deployment-options
  • profiles-and-environments-multi-env-presets
  • recurring-task-cancelable
  • recurring-task-self-schedule
  • redeploy-agents
  • rollback-deployment
  • snapshot-recovery
  • troubleshoot-build-force-rebuild
  • troubleshoot-manifest-trace
  • view-agent-files
  • wait-for-external-input

Modified Scenarios — 30

  • add-config-cli
  • add-config-manifest
  • add-cors
  • add-env-vars-cli
  • add-env-vars-manifest
  • add-http-endpoint
  • add-initial-files-basic
  • add-secret-runtime
  • add-secret
  • add-simple-agent
  • add-third-party-dependency
  • annotate-agent
  • atomic-block
  • building-agents
  • change-durability-mode
  • configure-api-domain
  • create-a-new-project
  • http-endpoint-auth
  • http-endpoint-parameters
  • invoke-agent
  • make-http-request
  • rpc-1-invoke-and-await
  • rpc-2-trigger
  • rpc-3-schedule-future-call
  • rpc-4-phantom-agents
  • schedule-agent
  • stateless-agent
  • transactions-1-fallible-rollback-http-ledger
  • trigger-agent
  • view-agent-logs

Golem improvements / fixes

  • CLI MoonBit support: New agent_id_display module with MoonBit parsing/rendering
  • MoonBit SDK overrides

Changes to SDKs

MoonBit SDK (~130 files)

  • SDK tools updated to latest MoonBit
  • New SDK modules: filesystem, ffi, http, api/retry, rpc, agent guest/snapshot exports
  • WIT bindings: Extensive new/updated bindings across:
    • golem/agent, golem/api/oplog, golem/api/retry, golem/quota
    • golem/rdbms/* (ignite2, mysql, postgres, types)
    • wasi/filesystem/*, wasi/http/*, wasi/keyvalue/*, wasi/blobstore/*
    • and more

Scala SDK (33 files)

  • Cancelable scheduled invocations — new CancellationToken type + scheduleCancelableAt codegen
  • PrincipalConverter rewrite — hand-written JSON replacing derived JsonCodecDeriver
  • RDBMS numeric type fixes — number vs bigint handling for unsigned ints, oid, ordinal, and execute() returns
  • GolemPlugin simplification — removed unnecessary stale-linker-output detection
  • Build/config — .jvmopts, cross-build test alias fix

TypeScript SDK (3 files)

  • Cancelable scheduled invocations
  • Principal serialization aligned wit the other SDKs (it was broken due to bigint json serialization)

CI

  • The skill testing is now a daily job because it's too expensive to run it more frequently
  • It merges the reports and uploads a HTML report to the benchmark-results repo

@vigoo vigoo marked this pull request as ready for review April 21, 2026 19:23
@vigoo
Copy link
Copy Markdown
Contributor Author

vigoo commented Apr 22, 2026

Merging this with two IT failures that look like the known flaky issues, being investigated separately (so the coding agent test runs can begin)

@vigoo vigoo merged commit be19108 into main Apr 22, 2026
45 of 47 checks passed
@vigoo vigoo deleted the skill-harness-5 branch April 22, 2026 08:02
@github-actions github-actions Bot locked and limited conversation to collaborators Apr 22, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants