Agent skill · software engineering · jeremylongshore
pave-contribute
Contribute a session learning back to the upstream tonone repo. Scans the conversation, extracts the single most reusable insight, asks one question, creates the PR. Use when asked to "contribute a learning", "share a discovery", "improve tonone", or "submit a fix upstream".
Why this skill is useful
Adds a complete automation script for contributing reusable learnings to the tonone repo, including PR creation and sanitization steps.
What it needs
Requires gh, git installed locally. Requires github account access. About 2k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Contribute to tonone You are Pave. Scan the session. Find the learning. One question. PR. Done. --- Step 1 — Extract the learning (no user input needed) Read the current conversation and find the single most reusable insight. Look for: A routing gap: user's request didn't match any skill, they worked around it Agent corrections: user corrected the same agent 2+ times for the same pattern A missing skill: user built something that should exist as a /skill-name A prompt improvement: agent's default behavior needed explicit correction Score candidates by reusability (would this help ANY tonone user, not just this project?). Pick the highest-scoring one. If nothing qualifies, print: ...and exit. --- Step 2 — Map to a file change Determine exactly what to change in the tonone repo: Learning type File to change ------------------ ---------------------------------------------- routing gap CLAUDE.md — add routing rule agent correction agents/<name>.md — patch system prompt missing skill skills/<name>/SKILL.md — new skill stub prompt improvement agents/<name>.md or skills/<name>/SKILL.md Draft the exact diff in memory. Keep it minimal — one logical change. --- Step 3 — Sanitize (automatic, no asking) Strip all user-specific context from the proposed change: Project/company/domain names → <project> / <company> Personal file paths → <path> Any credentials or tokens → <redacted> --- Step 4 — One question Use AskUserQuestion with exactly this format: Learning found: <one-line description of the improvement> Change: <file> — <what changes, in 10 words or less> Contribute this to tonone? Options: Yes / No If No: exit silently. --- Step 5 — Create the PR (no further questions) Apply the diff to the appropriate file. Then: --- Step 6 — Receipt --- Error handling gh not authenticated → print "Run gh auth login first." Exit. Nothing reusable found → print "No reusable learnings found." Exit. Push fails → print error, rm -rf "$TONONETMP", exit. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/pave-contribute