Enqueue work items as candidate sets and activate them for validation.
npx skills add https://github.com/ether-moon/agent-atelier --skill candidate
Please help me install this Agent Skill. Skill: candidate Source: https://github.com/ether-moon/agent-atelier/tree/main/plugins/agent-atelier/skills/candidate Command: npx skills add https://github.com/ether-moon/agent-atelier --skill candidate
Candidates bridge implementation and validation. When a builder finishes work items, they are enqueued as a candidate set — one or more WIs sharing a branch and commit. The orchestrator activates the next set into the exclusive validation slot (one at a time). After validation, the set is cleared as completed or demoted (fate-sharing: all WIs succeed or all return to rework).
Enqueue WI-014 after implementing it to place the item into the candidate queue.
High-intent search queries and answers about installing and using this skill.
Batch promotion in validation pipelines queues WIs as a candidate set, activates them one at a time, and clears them with transactional updates. Fate-sharing ensures all WIs succeed or return to rework together.
Fate-sharing in candidate validation means all WIs in a batch succeed or fail together. When clearing the validation slot, the entire set is marked completed or demoted to rework, enforcing atomic batch consistency.
To enqueue multiple WIs for validation, provide comma-separated WI IDs. The orchestrator groups them into a candidate set sharing a branch and commit, then activates the set for validation using FIFO ordering.
You cannot activate multiple WI sets simultaneously. The pipeline enforces a single active validation slot, ensuring only one candidate set is validated at a time while others wait in the queue.
When validation fails, the candidate set is demoted back to rework. Due to fate-sharing, all WIs in the batch return together, and the validation slot is cleared so the next queued set can be activated.
novadbGenerate C# source code for multiple or all object types in a NovaDB branch.
grp06Analyzes a candidate shortlist and writes a decision.md with the chosen refactor.
anukkrit149Build LangChain agents with create_agent(), middleware, and structured outputs.
danger-EnglishDispatch independent tasks to specialized agents for parallel execution without shared state.
