sandbaseai/workbuddy-skill / skills/bug-triage
bug-triage
VerifiedTriage bug reports, analyze errors and stack traces, identify competing causes, find duplicates, and prioritize safe fixes with evidence.
Bug Triage and Fix Analysis
Use this skill when turning a bug report, exception, stack trace, failing test, or support signal into a bounded investigation and an owned next action. It helps distinguish a reproducible defect from a usage question, environment issue, duplicate, regression, or insufficient evidence. It does not claim to have accessed an issue tracker or repository unless that source was actually provided or authorized.
Intake and evidence quality
Capture the report ID or safe local reference, reporter context, first and most recent occurrence, time zone, affected version, environment, user-visible symptom, frequency, severity, workaround, and privacy classification. Preserve the original wording separately from normalized summaries. Mark each field as observed, reported, inferred, or unknown.
Request the smallest useful reproduction: exact input shape without secrets, expected and actual behavior, deterministic steps, a known-good comparison, relevant configuration, and the narrowest logs or trace. Redact tokens, credentials, personal data, private URLs, and confidential payloads before sharing or storing evidence. Do not ask a reporter to expose production data just to improve a ticket.
Analyze errors and stack traces
- Confirm the exception or failure is from the claimed component and version.
- Read the stack from the failure boundary outward, distinguishing the first application frame from framework or wrapper frames.
- Identify inputs, state transitions, dependency calls, and error handling at the boundary; do not treat the deepest or loudest frame as the root cause.
- Compare a failing and successful path, including deploy, configuration, schema, feature-flag, dependency, and traffic differences.
- Form multiple hypotheses and choose the cheapest safe experiment that separates them. Record what result would falsify each hypothesis.
- Confirm whether the behavior is a regression, an existing limitation, a duplicate, or a new failure mode before proposing a fix.
Never execute a suggested fix merely because it sounds plausible. Avoid broad production changes, deleting data, disabling validation, suppressing an error, or changing security controls as a debugging shortcut. Prefer read-only inspection, a minimized test, a local or controlled reproduction, and a small reversible change with an explicit rollback condition.
Triage and priority
Classify severity from demonstrated user or business impact, not from emotion or technical novelty. Consider affected scope, reproducibility, data loss or security exposure, workaround quality, duration, frequency, and whether impact is increasing. Keep severity, priority, confidence, and effort as separate fields:
- severity describes harm if the bug occurs;
- priority describes when the team should act under current constraints;
- confidence describes the quality of evidence;
- effort describes the estimated cost and uncertainty of the fix.
For duplicates, compare symptom, trigger, affected version, component, regression window, and proposed workaround. Link evidence rather than merging reports solely because their exception names match. If a security boundary or personal-data exposure may be involved, route it through the authorized security process and keep sensitive details out of ordinary issue comments.
Fix proposal and verification
A fix proposal must name the causal boundary, smallest safe change, compatibility impact, test that would fail before the fix, regression and adjacent boundary tests, rollout guard, rollback condition, and owner. Distinguish a mitigation from a durable fix. If the cause is not confirmed, state the next experiment instead of inventing certainty.
After implementation, verify the original reproduction, the relevant known-good path, input and permission boundaries, persistence or retry behavior, and the production signal that exposed the issue. For a rollout, compare affected and control segments over an appropriate observation window. Do not close a bug on the basis of a green build alone.
Handoff format
Bug reference / status / timestamp:
Observed symptom and expected behavior:
Scope, versions, environment, and user impact:
Evidence and redaction status:
Reproduction and known-good comparison:
Hypotheses with confidence and falsifying checks:
Duplicate / regression assessment:
Severity / priority / effort / rationale:
Mitigation or fix proposal / owner:
Verification, rollout, and rollback evidence:
Open unknowns and next update:
Return reproducible evidence and explicit uncertainty. A useful triage result ends with one safe next action and an accountable owner, even when the root cause remains unresolved.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |