onboarding: point the premise-probing doctrine at the probe skill - #62
Conversation
There was a problem hiding this comment.
Pull request overview
Updates the onboarding doctrine on premise-probing to point readers to a concrete, reusable procedure (the probe skill), reducing repeated re-derivation of how to run a probe in each session.
Changes:
- Adds a brief sentence in the “Probe the premise before building” section that directs users to the
probeskill as the procedural reference.
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
The section states the doctrine but not the procedure. The probe skill in agent-skills carries it: picking the load-bearing premise, refuting by arithmetic before commissioning a search, re-running prior art when the framing moves fields, and reporting what the probe could not reach.
|
Codex reviewed this alongside the two skills PRs it depends on, and raised the one thing that makes a global instruction file different from a skill: it is loaded by every session of every group member, including those who have not installed the group's Added a fallback: if the skill isn't installed (ONBOARDING step 5), the sentence itself is the checklist and gets run inline. The core doctrine above it was already self-contained, so nobody is left worse off than before. |
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.
Suppressed comments (1)
onboarding/CLAUDE.global.md:134
- This text assumes the
probeskill is available once skills are installed, but the referenced upstream change (agent-skills#62) may not have landed in the reader's localagent-skillscheckout yet. As written, someone who has installed skills could still not findprobeand be confused about what to do.
The procedure this doctrine implies lives in the `probe` skill: which premise to pick, trying to refute it by one line of arithmetic before commissioning a search, re-running prior art from scratch when the framing moves fields, and reporting what the probe could NOT reach alongside what it found. If the skill isn't installed (ONBOARDING step 5), that sentence is the checklist — run it inline rather than treating the procedure as unavailable.
CLAUDE.global.md's "Probe the premise before building" section states the doctrine but not the procedure, so each session re-derives how to actually run one.Adds two lines pointing at the new
probeskill (DanaResearchGroup/agent-skills#62), which carries the procedure: picking the load-bearing premise, trying to refute it by one line of arithmetic before commissioning a search, re-running prior art from scratch when the framing moves fields, and reporting what the probe could NOT reach alongside what it found.