hp-code-publish
claudePublishing work to GitHub: clear prose deliverables (write-clearly), outcome-led PRs (create-pr), and attribution headers on agent-authored comments (github-comment-attribution).
3 skills
AI code-review loop for PRs and branches: multi-lens review with HTML report, safe fixing, GitHub publishing, review-debt filing, a primitives-map generator, a verify-skill generator, and a class hardener.
Plugin hp-review-toolkit from marketplace humanity-plus in Humanity-Plus/commons. Commands read from the repo's manifest — 9 days ago.
create-verify-skillGenerate — or refresh — a repo-local `verify-<app>` skill that drives the real app the way a user does and proves a feature by its observable end state: launch, doctor, drive, evidence, cleanup, plus a feature map (one file per user-facing feature, with the source paths that implement it and the see
fix-the-classInvoke as /fix-the-class after the user corrects an agent for a mistake the repo has seen before ("again", "I keep telling agents", "make this impossible", "harden the repo"), or for a standing hardening pass over a repo's recurring agent mistakes. Finds the mistake classes from the repo's own evide
map-primitivesMap a product repo's system primitives into a canonical docs/primitives.md (+ machine-readable primitives.yaml) and detect primitive drift — the same concept re-declared in multiple shapes across schema, validators, API surfaces, and UI. Covers all seven primitive areas (data, API, auth, workflows,
publish-reviewPost the findings from a previous review-pr run to the GitHub PR as one real review — findings anchored in the diff become inline file:line comments, the rest folds into the review body. Always dry-runs first and confirms before posting. Use when asked to publish or post a review, send review findin
resolve-reviewSafely resolve the findings from a review — apply scoped fixes on your own trusted branch, validate them, and report exactly what changed and what still needs a human. Consumes a review-pr findings.json (or findings handed in-session), turns them into an ordered fix plan, fixes what's fixable (bugs,
review-fixReview the current branch (or your own PR) and fix it in one pass — runs the full review-pr pipeline in findings-only mode (no HTML report), then applies resolve-review's scoped fixes on the branch and reports a resolution table to chat. The fast local loop before opening a PR. Use when asked to rev
review-issuesFile a review's non-blocking findings (low/nitpick — readability, naming, soft conventions) as GitHub issues so they get fixed later instead of dying with the report: one issue per defect class (never per nitpick), deduped across runs, labeled review-debt for a scheduled fixer to drain. Use when ask
review-prMulti-lens code review for a pull request or local branch. Orchestrates independent review lenses (triage, first-five, conventions, warm, zombies, preflight, primitives, blast-radius) plus a PR-intent check, runs them as parallel subagents (sharded by feature area on large diffs, with a dedicated se
Publishing work to GitHub: clear prose deliverables (write-clearly), outcome-led PRs (create-pr), and attribution headers on agent-authored comments (github-comment-attribution).
3 skills
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.