القائمة
Git workflow for AI-assisted advocacy development — atomic commits per subtask, ephemeral branches, PR curation into reviewable chunks, AI-Assisted tagging
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Security audit workflow for advocacy projects — dependency verification, zero-retention compliance, slopsquatting defense, encrypted storage, instruction file integrity, device seizure readiness, ag-gag exposure assessment
Structured stakeholder interview for advocacy projects — gathers purpose, threat model, coalition needs, legal exposure, user safety requirements, and budget constraints one question at a time
Plan-before-code workflow — read existing code, write spec, decompose into subtasks, implement and test one at a time, comprehension check before committing
SEO + GEO audit and implementation workflow — Core Web Vitals, HTML structure, semantic writing, E-E-A-T, content intent, Wikipedia/Wikidata, JSON-LD schema, meta tags, crawl budget, robots.txt, sitemap, IndexNow, topic cluster architecture, link building, brand signals, conversion optimization, analytics, internationalization, platform presence, defensive review
Spec-first test generation, assertion quality review, mutation testing, five anti-patterns to avoid — for AI-assisted advocacy development where silent test failures mean lost evidence or exposed activists
Layered code review pipeline — automated checks first, then AI-assisted review, then human review focused on Ousterhout red flags, AI failure patterns, silent failures, and advocacy-specific concerns
| name | git-workflow |
| description | Git workflow for AI-assisted advocacy development — atomic commits per subtask, ephemeral branches, PR curation into reviewable chunks, AI-Assisted tagging |
Create a short-lived branch for the current task. Trunk-based development remains the goal — this branch is a safety net, not a long-lived workspace. If the agent has not produced mergeable work within one session, delete the branch and reconsider the approach. Name the branch for the task, not for a feature epic.
Break the overall task into the smallest logical subtasks. Each subtask is one commit. If the task decomposes into "extract interface, implement adapter, update callers," those are three separate commits. Never let the agent complete an entire multi-step task before committing.
Run the relevant test subset before each commit. Every commit must leave the codebase in a passing state. If tests fail, fix the issue before committing — do not push broken commits to shared branches. For advocacy code handling investigation or evidence data, also verify that no sensitive data has leaked into test output, logs, or error messages.
Write commit messages that explain WHY, not WHAT — the code shows what changed. First line: 50 characters, imperative mood. Reference the issue or ticket. Add appropriate AI attribution trailers so the team knows which code was agent-generated.
Continue the cycle: implement one subtask, test, commit. Each commit should be independently understandable. If you read the commit in isolation six months from now, you should know what it does and why.
PR curation is the critical human skill in AI-assisted development. AI adoption inflated PR size by 154%. Do not submit the agent's full output as one PR. Split into reviewable chunks:
Squash-merge ephemeral branches to keep trunk history clean. Delete branches immediately after merge. If a branch has lived longer than one working session, evaluate whether the approach needs to change rather than extending the branch's life.
