The debugging workflow skill in obra/superpowers: When encountering bugs, test failures, or abnormal behavior, it mandates that the agent first investigate the root cause before attempting fixes, following four stages: "Root Cause Investigation → Pattern Identification → Hypothesis and Verification → Implement Fix." The agent is not allowed to proceed to the next stage until the previous one is completed.
skills.sh install count 312061 (+1367 / 24h)
Security Audit: Not audited
- Not yet scanned
Rationale
- In the first stage, read error messages and stack traces, ensure stable reproduction, and check recent changes; for multi-component systems (e.g., CI → Build → Signing), add logging at each boundary first, run once, and identify at which layer the failure occurs.
- In the third stage, propose only one hypothesis at a time and change only one thing to verify it; before fixing, you must first write a failing test that reproduces the issue.
- If the issue persists after 3 attempts, stop, question the architecture, and discuss with the user instead of trying a 4th fix; also list thoughts like "patch it now and investigate later," which require returning to the first stage immediately upon arising.
- skills.sh installs: 1367 in 24 hours, cumulative 312061; the same repository contains 14 other skills (indicating attention only).
Usage
- Automatically triggered when encountering bugs, test failures, build failures, etc., without requiring additional commands.
- The fix stage invokes the test-driven-development and verification-before-completion skills from the same repository; it is best to install them together. Requires an agent environment capable of running tests and commands.
Risks
- Not audited in this period, fixed at C-level, and install commands are not displayed; SKILL.md is purely procedural documentation, with diagnostic commands in examples such as printing whether environment variables exist and listing macOS Keychain identities, provided for demonstration only.
- Auxiliary files in the same directory, such as root-cause-tracing.md and defense-in-depth.md, are outside the scope of this review.
- Note the distinction from cloned accounts like skills-shell/superpowers, which are automatically excluded from the leaderboard; this entry comes from obra/superpowers. The workflow is heavy, requiring all four stages even for minor issues, which increases time and token usage.