adr
Write an Architecture Decision Record (ADR) for a feature — Context / Decision / Status / Consequences / Alternatives, filed as…
Uncommitted code risk assessment with breaking change detection, blast radius analysis, and scope metrics. Use when: evaluating PR risk, pre-commit risk check, large refactoring review. Not for: security vulnerabilities (use /codex-security), code correctness (use
Claude Code
$ npx -y skills add sd0xdev/sd0x-harness --skill risk-assess --agent claude-code
Codex
$ npx skills add sd0xdev/sd0x-harness --skill risk-assess -a codex -y
Cursor
$ npx skills add sd0xdev/sd0x-harness --skill risk-assess -a cursor -y
Gemini CLI
$ npx skills add sd0xdev/sd0x-harness --skill risk-assess -a gemini-cli -y
GitHub Copilot
$ npx skills add sd0xdev/sd0x-harness --skill risk-assess -a github-copilot -y
OpenCode
$ npx skills add sd0xdev/sd0x-harness --skill risk-assess -a opencode -y
Windsurf
$ npx skills add sd0xdev/sd0x-harness --skill risk-assess -a windsurf -y
name: risk-assess description: "Uncommitted code risk assessment with breaking change detection, blast radius analysis, and scope metrics. Use when: evaluating PR risk, pre-commit risk check, large refactoring review. Not for: security vulnerabilities (use /codex-security), code correctness (use /codex-review-fast). Output: 3-dimension weighted score + risk level + gate."
1. Run `bash scripts/run-skill.sh risk-assess risk-analyze.js --json` to collect deterministic scores 2. Parse the JSON output — overall_score, risk_level, dimensions, flags, gate, next_actions 3. **If risk_level = Critical** (score 75-100) — highlight all breaking signals, recommend splitting PRs 4. **If risk_level = High** (score 50-74) — auto-escalate to `--mode deep`, detail blast radius 5. **If risk_level = Medium** (score 30-49) — summarize dimensions, note areas of concern 6. **If risk_level = Low** (score 0-29) — brief summary, confirm safe to proceed 7. Add qualitative interpretation beyond the scores (e.g., "high blast radius but all dependents are test files")
The script analyzes 3 dimensions + 2 conditional flags:
| Dimension | Weight | What It Measures | |-----------|--------|-----------------| | breaking_surface | 45% | Removed exports, renamed APIs, changed signatures, deleted modules | | blast_radius | 35% | Number of files importing changed modules (grep-based) | | change_scope | 20% | File count, LOC delta, directory span, rename ratio |
| Flag | Trigger | What It Checks | |------|---------|---------------| | migration_safety | Migration/schema files in diff | Rollback/down file exists | | regression_hint | (v2 stub) | Future: git history analysis |
| Score | Level | Gate | Exit Code | |-------|-------|------|-----------| | 0-29 | Low | PASS | 0 | | 30-49 | Medium | PASS | 0 | | 50-74 | High | REVIEW | 1 | | 75-100 | Critical | BLOCK | 2 |
If the script fails, report the error and suggest running manually:
bash scripts/run-skill.sh risk-assess risk-analyze.js --json
## Risk Assessment Report | Field | Value | |-------|-------| | Score | **[N]/100** | | Risk Level | [icon] [level] | | Gate | [PASS/REVIEW/BLOCK] | ### Dimensions [table of dimension scores + weights] ### Breaking Change Signals [list of detected signals — only if any] ### Next Actions [prioritized action items] ## Gate: [sentinel]
Language: English | 繁體中文 | 简体中文 | 日本語 | 한국어 | Español The harness layer for Claude Code. Let the model choose the path. Keep "done" verifiable. Claude has discretion inside a closed, test-pinned anchor set.
Write an Architecture Decision Record (ADR) for a feature — Context / Decision / Status / Consequences / Alternatives, filed as…
Install and set up the optional agentctl mod (task scope, refusal before execution, evidence that goes stale, model-free…
Architecture design and documentation. Produces 3-architecture.md with component diagrams, data flow, integration points, and…
Context-aware Q&A with auto context gathering. Use when: user has a quick question about codebase, git history, rules, docs, or…
Industry best practices conformance audit with mandatory adversarial debate. Produces audit artifact: verdict (OK/WARN/FAIL) +…
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or…
Ask which skill or flow fits your situation. A router over the skills in this repo.