Auto tinker review
Local-first, chat-operated developer experiments, learning history, and approval-gated publishing.
npx -y skills add NolanCassidy/auto-tinker --skill auto-tinker-reviewAssembled 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
Summarize evidence-backed Auto-Tinker status and inspect the local graph, main-goal progress, queue health, and writing readiness. Use for daily, weekly, project, or capability reviews and questions about what was done, learned, blocked, scheduled, or connected; this skill does not execute code or mutate GitHub.
SKILL.md
2.8 KB, 529 tokens by cl100k_base, as published. Nobody here has run it
Auto-Tinker Review
Review canonical Markdown and exact evidence; do not turn missing repository links into missing knowledge.
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
- For a daily closeout, run
auto-tinker history reconcile --workspace <path> --jsonbefore summarizing. - Run
auto-tinker index, thenauto-tinker goal show,auto-tinker graph,auto-tinker queue list, andauto-tinker discoverwith--json. - Trace claims to event, evidence snapshot, experiment, test, lesson, journal, changelog, or source record IDs.
- Classify repository locations as verified local, verified remote, both, missing, or unverified. Preserve and report knowledge nodes that have no repository.
- Assess each active/finished item against the one active main goal: contribution, distraction cost, success criteria, and exploration budget. Keep supporting goals separate.
- Separate observed facts from inference and unanswered questions. Do not use activity volume as capability growth.
- Review an individual writing artifact only on user direction with
auto-tinker journal review <journal-id> --state <pending|approved>. This changes writing-review state only. - Generate the requested next action with
auto-tinker prompt <intent> [--target <id>] [--agent <name>]; present it as copyable text for a chat agent.
Read review-cadences.md for daily, weekly, and graph-query output contracts.
Boundaries
- A review is read/reconcile/index work. It does not start coding or publish remotely.
- Report
private-journal,readme,changelog, andpublic-storyas separate linked artifacts with separate readiness. - Report
readme_review,public_story_review, andrepository_publication_approvalseparately from one another and from actual remote visibility. - Never treat
journal review,writing_approval, README review, or public-story review as consent to make a repository public. - Flag stale machine/profile data, weak evidence, unresolved duplicates, broken links, blocked work, main-goal drift, and privacy risks.
- The viewer may display these results and edit safe local review/queue state, but it never invokes this skill itself.
Return exact record paths/IDs, evidence strength, changed reconciliation counts, main-goal progress, next decisions, and copyable prompts.
What ships with it: 2 files
2.4 KB alongside SKILL.md
agents/
- openai.yaml205 B
references/
- review-cadences.md2.2 KB