Auto tinker queue
Local-first, chat-operated developer experiments, learning history, and approval-gated publishing.
npx -y skills add NolanCassidy/auto-tinker --skill auto-tinker-queueAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 18 days oldThe repository was created 18 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Change local Auto-Tinker candidate and queue state by listing, starring, ranking, scheduling, grouping, blocking, deferring, skipping, or selecting items. Use for ordering or picking work only; this skill does not execute experiments.
SKILL.md
2.2 KB, 434 tokens by cl100k_base, as published. Nobody here has run it
Auto-Tinker Queue
Keep the human in control of what runs and why.
Invoke the CLI as auto-tinker; from an unlinked source checkout use npm --prefix <auto-tinker-product-repo> run cli -- followed by the same arguments.
Workflow
- Run
auto-tinker goal showandauto-tinker queue list --workspace <path> --json; preserve returned goal/candidate/queue IDs. - Translate conversational changes into explicit
auto-tinker queue update <id>calls using--starred,--priority,--rank,--scheduled-for,--blocked-reason,--goal,--goal-contribution,--distraction-risk, or--status. - Read the queue again and verify exact resulting order/state.
- For selection, run
auto-tinker queue next --count <n> --workspace <path> --json. Never substitute the highest discovery score for this command's policy-aware result. - Present selected scope, time, prerequisites, machine fit, privacy, main-goal contribution, distraction risk, and a copyable
$auto-tinker-runprompt. Stop without executing it.
Read queue-policy.md for states, precedence, and batch selection.
Rules
- Manual rank, pin/star intent, block, and explicit schedule beat automated scores.
- Never silently rewrite the user's goal to improve ranking.
- Keep exactly one active main goal. A manually selected distraction is allowed, but label its tradeoff instead of pretending it advances the goal.
- Do not select blocked, skipped, archived, duplicate, incompatible, or unreviewed-risk candidates.
- A queue selection changes local state only. It does not authorize planning, code execution, GitHub mutation, or unrelated filesystem changes.
- Starting multiple items does not override configured concurrency or resource limits.
- Keep an actionable
goaland next action on ready work; send unclear items back for refinement.
Return the final ordered list, changes made, selected IDs, exclusions with reasons, and a copyable run prompt.
What ships with it: 2 files
1.8 KB alongside SKILL.md
agents/
- openai.yaml198 B
references/
- queue-policy.md1.6 KB