onboard-light
Skill1 starsChanged 57 days agoMITView on GitHub
Onboard Claude onto a repo without touching .claude/: the full onboard-repo flow (recon, mutual interrogation, verified feedback loop, CLAUDE.md block, wiki, cold eval) with the harness phase (4b) skipped by standing instruction. Use when the user says "onboard light", "onboard without the harness", "onboard but no hooks/settings/rules", or wants knowledge-only onboarding on a repo where enforcement config is unwanted (a client repo, a restricted repo, or one that is not theirs to configure). For the full flow including the .claude/ harness use onboard-repo; to add only the harness to an already-onboarded repo, run onboard-repo and re-run phase 4b alone.
In this repository64 other files it gives its agents.
- accioSkill
- anthropic-doc-validatorSkill
- archifySkill
- cavecrewSkill
- caveman-compressSkill
- caveman-exploreSkill
- caveman-helpSkill
- cavemanSkill
- codebase-designSkill
- conference-talk-deckSkill
- config-gcSkill
- context-budgetSkill
- domain-modelingSkill
- excalidrawSkill
- explain-yogi-like-he-is-5Skill
- field-guide-builderSkill
- gateguardSkill
- geo-contentSkill
- goal-creatorSkill
- grillingSkill
- grill-meSkill
- grill-with-docsSkill
- hindsightSkill
- implementSkill
- improve-codebase-architectureSkill
- loop-design-checkSkill
- mental-modelsSkill
- migrationSkill
- modern-pythonSkill
- no-ai-slopSkill
- onboard-repoSkill
- ponytail-auditSkill
- ponytail-reviewSkill
- ponytailSkill
- primitive-classifierSkill
- prompt-fixerSkill
- prompt-optimizerSkill
- prototypeSkill
- qa-deckSkill
- read-wikiSkill
- researchSkill
- resolving-merge-conflictsSkill
- safe-refactorSkill
- scope-creep-checkSkill
- sdd-workshop-walkthroughSkill
- setup-matt-pocock-skillsSkill
- skill-complySkill
- skill-scoutSkill
- smellsSkill
- spec-interviewSkill
- surgical-patchSkill
- teachSkill
- the-llm-councilSkill
- to-questionnaireSkill
- to-specSkill
- to-ticketsSkill
- triageSkill
- wait-whatSkill
- watchSkill
- wayfinderSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Treat the Phase 4b checkpoint as already answered: "later": Do not run the harness interview, do not write anything under .claude/, and do …
- Very specific: line 41
If, mid-run, the user asks for hooks, settings, rules, or anything under .claude/, that is no longer light: say so, and continue under …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
onboard-light
A standing pre-answer over onboard-repo, not a second process. The entire method lives in
~/.claude/skills/onboard-repo/SKILL.md; this wrapper changes exactly one decision, so the two
can never drift apart.
What to do
- Invoke the
onboard-reposkill and follow it in full: same phases, same checkpoint protocol, same references. - Treat the Phase 4b checkpoint as already answered: "later". Do not run the harness
interview, do not write anything under
.claude/, and do not ask; the choice was made by invoking this skill. - Record the skip in
knowledgebase/wiki/repo-onboarding.mdas "harness unbuilt, skipped via onboard-light", so a future run sees the gap. The existing edge case ("wiki exists, no.claude/HARNESS.md: recommend Phase 4b alone") then offers the harness whenever the user wants it, with Phases 0 to 3 read from the wiki instead of re-derived. - Everything else is unchanged: every checkpoint still stops and waits, Phase 3 still executes commands only with approval, Phase 5 still runs cold.
At the close, say in one line that the harness can be added any time by running onboard-repo and
choosing to re-run phase 4b alone. Do not sell it harder than that. On a Python repo with the
modern-python skill installed, add one more line: that skill offers its own interview on the
first Python edit, and a modern-python: off line in the CLAUDE.md block keeps it advisory
without writing anything under .claude/.
Boundary
If, mid-run, the user asks for hooks, settings, rules, or anything under .claude/, that is no
longer light: say so, and continue under plain onboard-repo rules (run Phase 4b normally at its
checkpoint) rather than improvising harness writes inside the light flow.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.