| name | studyos-quality-review |
| description | Review StudyOS code changes for security, privacy, reliability, dead code, duplicate logic, regressions, missing validation, and maintainability. Use for PR reviews, pre-merge checks, "review this", "security review", "quality review", cleanup audits, or before claiming substantial StudyOS implementation work is ready. |
StudyOS Quality Review
Overview
Use this skill to review StudyOS changes like a senior engineer. Lead with
actionable findings, prioritize real risk, and verify with the narrowest useful
commands.
Review Stance
Start from the diff and surrounding contracts:
- Identify the changed files and any existing unrelated worktree changes.
- Read nearby code before judging intent or style.
- Check tests, docs, config, deployment paths, and runtime prompts when behavior
crosses those boundaries.
- Prefer concrete bugs over broad style opinions.
- Do not invent issues. If a concern is speculative, label it as such and say
what would verify it.
Risk Checklist
For each changed area, check:
- Security and privacy: tokens, webhook secrets, Discord/GitHub IDs, auth
profile routing, command execution, file upload roots, logs, and prompts.
- Reliability: explicit errors, retry behavior, race conditions, persistence,
deployment volume behavior, and restart safety.
- Behavior: broken API contracts, changed prompt semantics, incorrect Discord
thread/channel routing, artifact upload mistakes, or GitHub actor confusion.
- Code health: dead branches, unused helpers, duplicate logic, over-large files,
unclear boundaries, hidden fallback behavior, and mock production data.
- Validation: missing tests for changed behavior, stale docs, or commands that
should have been run before handoff.
Review Procedure
- Inspect
git status --short and separate unrelated changes from the review
target.
- Inspect the relevant diff with line context.
- Read the local implementation, tests, docs, and config that define the
touched behavior.
- Run focused checks when practical. For this gateway, prefer
ruff check .,
pyright, and pytest; for target repos, use their own documented checks.
- Report findings first, ordered by severity, with file and line references.
- Include open questions only when an answer changes the recommendation.
- Keep summary secondary and short.
Finding Format
Use this shape:
- Severity:
Critical, High, Medium, or Low.
- Reference: clickable file path with the most relevant line.
- Issue: what can go wrong and under which condition.
- Fix direction: the smallest change likely to address it.
If there are no findings, say that clearly and mention remaining test gaps or
residual risk.
Non-Goals
- Do not rewrite code during a pure review unless the user asked for fixes.
- Do not demand unrelated cleanup.
- Do not collapse separate concerns into one vague issue.
- Do not expose secrets or personal data in review output.