Add Claude Code GitHub Workflow#3
Merged
Merged
Conversation
6 tasks
5 tasks
5 tasks
cailmdaley
added a commit
that referenced
this pull request
May 13, 2026
…between Restructures INTERVIEW to run the paper-extraction substrate inline, between two question beats. This way every interview question after the paper identifier is grounded in the actual paper — scope shown as a menu of real figures/tables, fidelity intent referencing actual headline numbers, code repo confirmed against what data-availability / acknowledgements surface, conventions/warnings proactively proposed from the paper, "prior notes" reframed as "external context outside paper + code." The paper identifier is asked in prose (not AskUserQuestion — the answer is inherently free-form: arXiv ID, DOI, or PDF path). AskUserQuestion lives entirely in Beat 2, after paper substrate is on disk. ACQUIRE becomes thin: just /lc-from-code scan-only against the cloned reference repo, plus code-status.yaml. Paper substrate is INTERVIEW's deliverable (committed as part of INTERVIEW's first commit, alongside constitution.md + CLAUDE.md). No-public-repo case is one step: write code-status.yaml found=false and launch. Knock-on doc edits: - SKILL.md Phases table: paper substrate moves to phase 0 outputs. - SKILL.md "The two pre-loop bookends": INTERVIEW now describes the two-beat shape; ACQUIRE drops the "two parallel sub-skill" framing. - SKILL.md Workdir-as-state: INTERVIEW row now includes paper substrate; ACQUIRE row simplified. - SKILL.md Resuming: split the substrate-incomplete recovery into paper-side (re-run /paper-extraction) and code-side (re-run /lc-from-code) cases. - references/acquire.md: full rewrite — drops paper-extraction step, becomes code-substrate-only with a "no-public-repo" branch. - references/architect.md: survey signal reworded to reference paper substrate (from INTERVIEW) and code substrate (from ACQUIRE) separately. Surfaced via Cail's dogfood run of /lc-from-paper on arXiv:2604.03227. Friction entries #2 (AskUserQuestion is wrong shape for free-form paper ID), #3 (INTERVIEW should ground questions in actual paper), and #6 (prior-notes framing doesn't match either scientist persona) all addressed here. Co-Authored-By: Claude Sonnet 4.7 (1M context) <noreply@anthropic.com>
cailmdaley
added a commit
that referenced
this pull request
May 13, 2026
Cail's call after walking through dogfood friction #3: instead of two pre-loop phases (INTERVIEW asks → ACQUIRE clones + scans), one ORIENT phase that runs the right thing at the right moment. ORIENT runs as one phase with seven stages, all in the user's main session: 1. Ask for the paper (prose). 2. /paper-extraction inline; read the substrate. 3. Interview the user (AskUserQuestion rounds, grounded in paper). 4. Clone repo + /lc-from-code scan-only (skip if no public code). 5. Optional follow-up questions if the code-index surfaced anything. 6. Draft constitution.md + CLAUDE.md (now informed by paper AND code). 7. User review → refine → single commit (constitution + CLAUDE + paper substrate + code substrate) → launch loop. The key shape: knowing the code BEFORE writing the constitution lets the constitution's Scope and sub-analysis decomposition lean on the actual pipeline. Previously the constitution drafted from paper-only context, then ACQUIRE added the code substrate after. Cleaner this way. Knock-on edits: - SKILL.md: phases table goes 0 ORIENT → 7 REVIEW (eight phases, was nine). "Two pre-loop bookends" section collapses into "The pre-loop bookend: ORIENT". Workdir-as-state's INTERVIEW + ACQUIRE rows merge. Resuming + anti-patterns updated. - references/orient.md: new file (content carried from interview.md via git mv in the prior commit, fully rewritten here with the seven stages, Stage 4's code-substrate work folded in from the deleted acquire.md, Cail's "Author / worked closely with authors" replacing the prior "Coauthor / ..." option in familiarity). - references/architect.md: survey signal reworded for ORIENT. - references/review.md: "first being INTERVIEW" → "first being ORIENT". - templates/constitution.md: "INTERVIEW" / "ACQUIRE" → "ORIENT". Friction #3 closes cleanly with this; #2 (prose paper-identifier) and #6 (reframed external-context question) were already folded into the prior commit and carry forward into the new orient.md. Co-Authored-By: Claude Sonnet 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
🤖 Installing Claude Code GitHub App
This PR adds a GitHub Actions workflow that enables Claude Code integration in our repository.
What is Claude Code?
Claude Code is an AI coding agent that can help with:
How it works
Once this PR is merged, we'll be able to interact with Claude by mentioning @claude in a pull request or issue comment.
Once the workflow is triggered, Claude will analyze the comment and surrounding context, and execute on the request in a GitHub action.
Important Notes
Security
There's more information in the Claude Code action repo.
After merging this PR, let's try mentioning @claude in a comment on any PR to get started!